The Authorities in Azad Jammu and Kashmir have shut down all educational institutions as floods triggered by heavy rains caused widespread destruction, reported 24NewsHD TV channel.
A notification in this regard was issued on Friday. According to which the educational institutions will remain closed on Friday and Saturday due to inclement weather.
Reporter Ishfaque Abbasi
Pakistan Refinery Limited (PRL) has announced a 15-day shutdown for regeneration, starting from August 17, 2025.
The refinery will remain non-operational during this period as part of its scheduled maintenance activities.
In a notification to the Pakistan Stock Exchange (PSX) on August 15, 2025, PRL stated that the shutdown is necessary to carry out essential regeneration processes.
PRL, incorporated in May 1960 as a public limited company, is engaged in the production and sale of petroleum products.
It operates as a subsidiary of Pakistan State Oil Company Limited (PSO) and plays a vital role in meeting the country’s energy needs by supplying refined petroleum products to the domestic market.
Gold headed for a weekly loss, after traders pared bets on the Federal Reserve cutting rates next month following a pick-up in inflation.
Bullion traded near $3,340 an ounce, after ending the previous session 0.6% lower following a report that showed US wholesale inflation accelerated in July by the most in three years. Bond yields and the dollar advanced after the data print, weighing on non-interest bearing gold as it is priced in the currency.
BEIJING, Aug. 15 — China’s consumer spending in the first seven months of this year expanded at a solid pace, with retail sales of consumer goods growing 4.8 percent year on year, official data showed Friday.
In July alone, retail sales of consumer goods rose 3.7 percent from a year earlier to nearly 3.88 trillion yuan (about 543.6 billion U.S. dollars), according to the National Bureau of Statistics (NBS).
The consumer goods trade-in program continued to drive growth. Among major retailers, sales of household appliances and audio-visual equipment surged 28.7 percent year on year, while furniture rose 20.6 percent, the data showed.
Online retail sales remained a bright spot, climbing 9.2 percent year on year in the first seven months. Sales of physical goods online increased 6.3 percent, making up nearly a quarter of total retail sales.
Service consumption also showed stable growth, with retail sales in the sector rising 5.2 percent year on year in the Jan.-July period. Retail sales of travel consulting and rental services, transportation services, and cultural, sports and leisure services all posted double-digit growth, NBS spokesperson Fu Linghui told a press conference.
Policy measures have sustained the expansion of consumption and fostered new growth drivers, Fu said, while warning that uncertainties abroad and constraints on domestic consumption capacity still pose challenges.
Looking ahead, efforts will be made to further implement targeted initiatives to boost consumption, cultivate new growth drivers in services, improve the consumption environment, and promote the stable development of the consumer market, Fu noted.
This review focuses on the integration of artificial intelligence (AI) within medical physiology education in China, encompassing three primary learner groups to reflect the diversity of medical training pathways. The scope includes: (1) undergraduate medical students (eg, clinical medicine, pediatrics, and preventive medicine), who form the core of China’s medical workforce; (2) allied health science students (eg, nursing, radiology, and biomedical engineering trainees), for whom physiological knowledge is critical to clinical practice; and (3) Traditional Chinese Medicine (TCM) practitioners undergoing modernized training, where AI is increasingly applied to bridge classical TCM theories (eg, meridian physiology) with evidence-based modern physiology. The analysis prioritizes core physiological domains aligned with China’s national medical education standards, including cardiovascular, respiratory, neurophysiological, and endocrine systems, to ensure relevance to clinical practice and policy directives such as the Medical Education Reform and Development Plan (2021–2025).
Historically, physiology education in China has been rooted in a structured, multi-modal framework designed to establish foundational knowledge and practical skills. Central to this approach are formal lectures, where instructors deliver theoretical content (eg, mechanisms of ion transport, endocrine regulation) to large groups, ensuring standardized dissemination of core concepts. Complementing lectures are small-group tutorial sessions, facilitating interactive discussions to clarify complex topics (eg, interpreting electrocardiograms, analyzing renal function parameters) and fostering critical thinking. Instructor-guided laboratory experiments further enhance learning, with hands-on training using animal models (eg, frog neuromuscular junction studies) or human simulators to practice physiological measurements (eg, blood pressure monitoring, spirometry). Finally, self-directed learning—primarily through canonical textbooks and supplementary materials—reinforces concepts outside class hours, encouraging independent knowledge consolidation. While effective in standardizing curricula, these traditional methods face challenges in addressing learner diversity (eg, varying prior knowledge) and adapting to emerging demands, such as integrating omics data or AI-driven tools into physiological education.
Within this review, artificial intelligence (AI) refers to computational systems capable of performing tasks historically requiring human intelligence, with a focus on two key categories driving advancements in modern medical education.1 Generative AI encompasses systems (eg, large language models [LLMs] like ChatGPT-4, image generators) that create novel content (text, images, or simulations) by learning patterns from training data. In physiology education, these tools enable dynamic question generation (eg, “Explain the role of the renin-angiotensin system in blood pressure regulation”) and interactive case-based learning, enhancing engagement and adaptability. Machine learning (ML), the second category, involves algorithms (eg, deep learning, reinforcement learning) that improve task performance (eg, diagnostic accuracy, simulation fidelity) through iterative data analysis. For instance, ML-driven physiological simulators (eg, virtual models of autonomic nervous system regulation) and adaptive learning platforms—which tailor content to individual student progress—are increasingly applied to refine training outcomes. Together, these AI subtypes are reshaping how physiological knowledge is taught, learned, and applied in Chinese medical education landscape.
Although studies from countries such as Canada have demonstrated high acceptance of AI among medical students, China’s medical education system is characterized by unique features of “strong policy guidance combined with regional resource disparities” (eg, the China’s Medical Education Reform and Development Plan (2021–2025) proposes the goal of “empowering primary medical education with AI”). Thus, a targeted analysis of adaptive pathways for AI integration in China’s medical physiology education is warranted.
The development of the physiology curriculum in China has been shaped by various factors. Curriculum design components, including vision, operationalization, design, and evaluation, interact dynamically, and are influenced by factors such as policy, local context, societal expectations, research trends, and technology.2 In the National University of Rwanda, a new modular curriculum in physiology received positive feedback regarding active learning methods, but faced challenges related to limited contact hours and resources.3
In China, Sichuan University West China Medical School’s undergraduate sonographer education program, which included an international curriculum on ultrasound physics and hemodynamics, showed comparable teaching effects between international remote teaching mode before the COVID-19 pandemic and domestic on-site teaching mode during the pandemic.4 Additionally, in a Brazilian public university, students considered human physiology and exercise physiology courses important yet challenging, suggesting the need for continuous assessment to improve the teaching-learning process.5
The integration of AI in Chinese medical institutions holds great potential but also faces several challenges. Ethical concerns are a key obstacle, as emphasized in the twelve tips for addressing ethical issues in the implementation of AI in medical education.6 These include the need to emphasize transparency, address bias, validate content, prioritize data protection, obtain informed consent, foster collaboration, train educators, empower students, regularly monitor, establish accountability, adhere to standard guidelines, and form an ethics committee.
A cross-sectional study in Pakistan explored the integration of AI into medical education, finding that while the majority of participants had a positive attitude towards it, there were concerns about over-reliance, ethical issues regarding privacy and confidentiality, and the need for institutional support.7 In China, the integration of AI in radiology education has been investigated, with a survey showing that although most respondents were not familiar with how AI is applied in radiology education, a substantial proportion were eager for its integration.8
In summary, the historical development of AI in medical education, the evolution of the physiology curriculum in China, and the integration of AI in Chinese medical institutions all contribute to the foundation of AI-based medical physiology education. The trends in medical education reform, the recognition of the importance of AI, and the efforts to address ethical and practical challenges set the stage for the further development of AI in this field. However, there is still a long way to go in terms of fully integrating AI into medical physiology education, ensuring its ethical use, and providing adequate educational opportunities for students and educators alike.
In recent years, the application of artificial intelligence (AI) in medical education has gradually increased, especially in Chinese medical schools. The introduction of AI technology has brought many new opportunities and challenges to medical education.
First, the application of AI in medical education can significantly enhance personalized learning and clinical training. AI-driven simulation technology offers realistic and immersive training opportunities, helping students prepare for complex clinical scenarios and fostering the development of interdisciplinary collaboration skills.9 However, the introduction of AI also presents ethical challenges, particularly the risk of skill regression due to over-reliance on AI, as well as the widening digital divide between educational institutions.9
Secondly, the application of AI in medical education also involves the development of professional ethics education in medicine. Research indicates that Chinese medical graduate students have varying views on integrating AI with professional ethics education. These views encompass both current development paths and predictions for future trends. The potential benefits of AI in professional ethics education could significantly enhance the quality and effectiveness of medical education, but it is also important to be aware of potential risks and to implement cautious supervision and management.10
Finally, the application of AI in medical education requires establishing a trust framework to ensure that the use of AI tools is both safe and beneficial. By drawing on the theory of trust decision-making in medical education, a structured approach can be developed to define the role and extent of AI involvement in relevant tasks. This trust framework helps to clearly identify the risks associated with AI usage and develop strategies to mitigate these risks.11
In short, the application of AI in Chinese medical schools has brought new perspectives and possibilities to medical education, but it is also necessary to find a balance between technological innovation and humanistic skills to ensure that AI can effectively enhance the quality and effectiveness of medical education.
While generative AI tools (eg, large language models like ChatGPT-4) have shown promise in enhancing interactive learning through dynamic question generation and case-based scenarios, their integration into physiology education also introduces notable challenges to assessment integrity. Recent studies highlight that over-reliance on these tools may compromise learning outcomes, as students might substitute critical thinking with passive generation of answers for assignments or assessments. For instance, Lucas et al12 reported in a systematic review that 38% of surveyed medical students admitted using LLMs to complete written tasks, raising concerns about the validity of traditional assessment methods in evaluating true comprehension of physiological concepts. Morjaria et al13 further demonstrated that ChatGPT-generated responses to short-answer questions in physiology exams achieved pass rates comparable to human students (67% vs 72%), yet lacked depth in reasoning—a critical skill for clinical decision-making. These findings underscore the need for educators to adapt assessment strategies, such as incorporating real-time oral defenses or scenario-based practical tasks, to mitigate the risk of superficial learning and ensure alignment with the core goal of fostering deep physiological understanding.
In physiology education, various AI tools and platforms are being utilized to enhance the learning experience. For example, in the context of hemodialysis cannulation skills training, simulators with advanced sensors and computing methods could be indispensable tools for standardized skills assessment and training.14 These simulators can quantify cannulation skill using sensor data, potentially improving end-stage kidney disease patient outcomes.
In a comparative study of AI platforms in plastic surgery education, ChatGPT-4.0 outperformed other platforms on the Plastic Surgery In-service Training Examination, reaching accuracy rates up to 79%.15 This indicates its potential as an educational tool in this specialty. Additionally, in nursing education, a scoping review found that AI-based tools generally received positive attitudes from students, highlighting the need to further study generative AI tools within this context.16 These tools can potentially offer personalized learning experiences, which are crucial in the complex field of physiology education.
Case studies of AI implementation in Chinese medical schools provide valuable insights into its practical application. In some Chinese medical schools, team-based learning (TBL) is gradually being integrated, but its application remains limited.17 Only about less than half of the schools used TBL in basic medicine or clerkship disciplines, and only 10% used it in both. Challenges such as public awareness, executive support, professional training, and resource sharing need to be addressed to facilitate its wider adoption.
In the context of anatomy education, some schools face a shortage of anatomical specimens, and body donation programs are being promoted.18 AI-based technologies could potentially complement these efforts by providing virtual anatomical models for teaching. Moreover, in the implementation of AI in orthopaedic research, while it has the potential to transform the field by improving disease diagnosis, clinical decision-making, and outcome prediction, it also faces challenges due to the inherent characteristics and barriers of the healthcare sector.19
China’s policy-supported regional pilot projects, such as the National Virtual Simulation Experiment Teaching Project for’ AI + Physiology Education,’ aim to address the imbalance in resources between urban and rural areas through technological innovation and the optimal allocation of educational resources. By introducing advanced artificial intelligence technology and providing virtual simulation platforms, these projects ensure that students in remote areas can access the same high-quality educational resources as those in cities. This approach not only enhances the accessibility and fairness of education but also helps to narrow the gap in educational resources between urban and rural areas.
Specifically, these pilot projects have broken the traditional reliance on physical laboratories and equipment through virtual simulation experiments, making the allocation of educational resources more flexible and efficient. Students can conduct experiments and learn through online platforms, greatly enhancing the convenience and flexibility of their learning experience. Furthermore, the application of artificial intelligence technology can provide personalized learning suggestions and feedback based on students’ progress and understanding, thereby improving learning outcomes.
This innovative educational model complements China’s new urbanization policy. This policy not only focuses on urban development but also emphasizes the growth of rural areas, aiming to reduce environmental pollution by optimizing energy structures and promoting green technological innovation.20 Similarly, pilot projects in the education sector are striving to achieve a balanced distribution of educational resources between urban and rural areas through technological means, promoting the joint development of both regions. These measures provide valuable experience and policy recommendations for China’s efforts to achieve ecological sustainability.
Student and faculty perceptions of AI in physiology education play a crucial role in its successful integration. A cross-sectional study of medical, dental, and veterinary students from 192 faculties worldwide showed that students had positive attitudes towards AI in healthcare (median: 4, IQR: 3–4) and a desire for more AI teaching (median: 4, IQR: 4–5), but had limited AI knowledge (median: 2, IQR: 2–2) and lack of AI courses (76.3%).21
In a study of Canadian medical students, 75% of respondents expressed confidence in giving presentations, but fewer were confident in providing bedside teaching, teaching sensitive issues, and presenting at journal clubs.22 A significant number (75%) also expressed interest in participating in a clinical teaching elective related to AI-enhanced teaching. These findings suggest that while there is interest in AI-related education, there is a need to address knowledge gaps and build confidence among students and faculty.
The current state of AI in medical physiology education in China shows that while there are promising AI tools and platforms, and some case studies demonstrate the potential of AI implementation, there are still challenges to overcome. Student and faculty perceptions highlight the need for increased AI knowledge, more AI-related courses, and confidence-building measures. Addressing these aspects will be essential for the successful and widespread integration of AI in medical physiology education.
Globally, there are significant differences in attitudes of medical students towards AI (AI) in medical education. Chinese medical students show some unique characteristics in AI education, which is in sharp contrast to the situation in other countries.
First, Chinese medical students show a high level of awareness and a positive attitude towards the application of AI in medical education. In one study, although only 43.5% of students were aware of the specific applications of AI in medical education, most students were familiar with the concept of AI.23 In contrast, medical students in other countries may have a lower level of awareness and application of AI. For example, in a global survey, only 38% of respondents had some knowledge of clinical AI, and the actual use of AI was only 20%.24
Second, Chinese medical students show a higher willingness and enthusiasm to participate in AI education. Research indicates that Chinese graduate students and male students are more inclined to integrate AI tools into their future learning and teaching activities.23 In contrast, while medical students in other countries are optimistic about the potential of AI in radiology, they are less concerned about AI replacing human radiologists and remain cautious about its practical applications.25
Finally, China also exhibits unique needs and challenges in the field of radiology AI education. Despite the positive attitude of Chinese radiologists towards AI-assisted medical imaging, more than half of the doctors still lack the necessary training.26 In contrast, while there is widespread interest and enthusiasm for AI in radiology education in other countries, trust and utilization of AI platforms need to be enhanced, particularly in user feedback systems and design.8,27
To sum up, there are significant differences between Chinese medical students and other countries in terms of AI attitude and radiology AI education. This difference not only reflects the uniqueness of China in the field of AI education, but also reveals the necessity of promoting AI education and application on a global scale.
Machine learning algorithms are making significant inroads in physiology simulations, offering new perspectives on understanding physiological processes and disease mechanisms. In a study comparing a deep learning algorithm to human gradings for detecting glaucoma on fundus photographs, the machine-to-machine (M2M) deep learning algorithm had a significantly stronger absolute correlation with standard automated perimetry global indices than human graders.28 The partial AUC for the M2M DL algorithm was also significantly higher, indicating its potential to replace human graders in population screening for glaucoma.
In another study, discrete multiphysics and reinforcement learning were combined to develop an in-silico model of human physiology, specifically focusing on the activity of the autonomic nervous system.29 Using the case of peristalsis in the oesophagus as a benchmark, the model effectively learned how to propel the bolus, demonstrating the power of combining first-principles modelling and machine learning in simulating physiological processes.
Moreover. Machine learning algorithms have increasingly become integral to the field of physiology simulations, offering innovative approaches to model complex biological systems. These algorithms provide the capability to analyze vast datasets, identify patterns, and make predictions that can enhance our understanding of physiological processes. The integration of machine learning with traditional physiology simulations has opened new avenues for research and clinical applications, allowing for more precise and personalized healthcare solutions.
One of the key areas where machine learning has shown promise is in the optimization of predictive models in perioperative medicine. Machine learning algorithms, such as decision trees and support vector machines, have been employed to improve the prediction of outcomes in perioperative settings. These models are compared with traditional statistical methods to assess their predictive performance. While machine learning models often show enhanced predictive capabilities, the clinical significance of these improvements remains a topic of discussion. The complexity and opacity of some machine learning models can pose challenges in clinical settings, where interpretability and transparency are crucial for decision-making.30
In the realm of laboratory medicine, machine learning algorithms are being developed to increase diagnostic accuracy and streamline operations. However, there is a growing awareness of the potential for these models to perpetuate or even amplify existing healthcare disparities if they learn from biased data. Efforts are being made to develop fair machine learning algorithms that mitigate these biases and promote equity in healthcare outcomes. This involves understanding what constitutes an unfair model and applying engineering principles to ensure fairness in algorithm development.31
Furthermore, the application of machine learning in mobile health (mHealth) for chronic disease management, such as asthma, demonstrates the potential of these algorithms to provide personalized feedback and improve patient outcomes. Machine learning techniques, including both supervised and unsupervised learning, have been utilized to analyze data from various devices like smartphones and smart inhalers. Despite the promising results, many studies face limitations due to small sample sizes and lack of external validation, which can hinder the generalizability of the findings. Future research is needed to address these challenges and validate the algorithms in real-world settings.32
Overall, the integration of machine learning algorithms in physiology simulations holds significant potential for advancing medical research and clinical practice. However, it is essential to address the challenges of model interpretability, fairness, and validation to fully realize the benefits of these technologies in healthcare.
Virtual reality (VR) and augmented reality (AR) applications are transforming medical physiology education by providing immersive and interactive learning experiences. In the field of congenital heart disease, advanced visualization techniques such as VR, AR, and 3D printing are being used to improve the understanding of complex morphologies.33 These technologies can help students and medical professionals better visualize the anatomical structures and physiological functions involved in the disease.
A review of VR, AR, and mixed reality (MR) applications in surgical simulation found that they can increase the fidelity, level of immersion, and overall experience of surgical simulators.33 For example, in maxillofacial surgery and neurosurgery, these technologies can provide more realistic training environments, allowing surgeons-in-training to practice complex procedures in a risk-free setting.
The integration of AI-driven virtual reality (VR) and augmented reality (AR) technologies in physiology education represents a significant advancement in medical training. These technologies offer immersive and interactive learning experiences that can enhance student engagement and facilitate the understanding of complex physiological concepts. The use of VR and AR in medical education is becoming increasingly prevalent, as they provide alternative methods to traditional teaching that can be particularly beneficial in disciplines such as physiology and anatomy.
A systematic review and meta-analysis have evaluated the impact of VR and AR on knowledge acquisition for students studying preclinical physiology and anatomy. The findings suggest that while there is no significant difference in knowledge scores when comparing VR and AR to traditional teaching methods, these technologies remain viable alternatives for education in health sciences and medical courses.34 This indicates that while VR and AR may not drastically improve test performance, they offer unique educational benefits that can complement traditional methods.
Moreover, the adoption of AR in medical education has been explored in various studies, highlighting its potential to create a highly stimulating learning environment. AR allows for the integration of digitally generated three-dimensional representations with real-world stimuli, which can simplify the delivery and enhance the comprehension of complex information. This is particularly relevant in the context of remote learning and interactive simulations, which have become more prominent due to the COVID-19 pandemic.35 The ability of AR to improve knowledge and understanding, practical skills, and social skills among medical students underscores its value as an educational tool.
In addition to these benefits, the use of VR and AR in medical education faces certain challenges, especially in resource-limited settings. Barriers such as infrastructure limitations, high costs, and lack of localized content can hinder the widespread adoption of these technologies. However, potential solutions, including cost-sharing models and teacher professional development, have been proposed to overcome these obstacles. By addressing these challenges, educators and policymakers can maximize the impact of immersive technologies in global education.36
Overall, AI-driven VR and AR technologies hold immense potential to transform physiology education by providing immersive and interactive learning experiences. While challenges remain, the continued exploration and integration of these technologies in medical education can lead to significant advancements in teaching and learning outcomes.
AI-driven personalized learning systems are tailored to meet the individual needs of students in medical physiology education. In an e-learning context for K-12 students, a personalized recommender system was proposed, which consists of four stages: student profiling, material collection, material filtering, and validation.37 This system aims to recommend materials that match the students’ learning styles and abilities, enhancing their learning experience.
In healthcare, blockchain-based federated learning has been proposed for categorizing healthcare monitoring devices.38 This approach can ensure data privacy while enabling the categorization of AI-of-Medical-Things (AIoMT) devices, which is crucial for personalized healthcare and medical education related to device-based physiological monitoring.
Technological advances in AI, including machine learning algorithms, VR and AR applications, and AI-driven personalized learning systems, offer great potential for enhancing medical physiology education. These technologies can improve the accuracy of physiological simulations, provide immersive learning experiences, and personalize education to meet the diverse needs of students. However, challenges such as data privacy in federated learning and the need for further validation of VR and AR in real-world medical education settings need to be addressed.
Artificial Intelligence (AI)-driven personalized learning systems are revolutionizing the educational landscape by offering tailored educational experiences that adapt to individual learning styles, preferences, and paces. These systems leverage advanced algorithms to analyze data from learners’ interactions, providing insights that help in customizing educational content and strategies. The integration of AI in education is not only enhancing learning outcomes but also addressing diverse educational needs, making learning more inclusive and effective.
One of the significant applications of AI in education is through adaptive learning systems, which are designed to adjust the difficulty and type of content based on the learner’s performance and engagement levels. This approach is particularly beneficial in fields such as rehabilitation science education, where AI-driven adaptive learning systems can enhance multimodal case-based learning. By synthesizing the applications of AI models, educators and policymakers can effectively incorporate AI into educational curricula, thereby improving the quality of education and training in healthcare and other domains.39
Moreover, AI optimization algorithms are playing a crucial role in higher education management and personalized teaching. These algorithms help solve complex educational management problems and enable personalized learning experiences. Empirical studies have shown significant improvements in student learning outcomes, engagement, and satisfaction when using AI-driven personalized teaching compared to traditional methods. However, challenges such as data privacy, algorithmic bias, and the need for human-AI interaction remain critical areas that need addressing to fully realize the potential of AI in education.40
In addition to these advancements, AI-driven personalized medicine is transforming clinical practice in fields such as inflammatory bowel disease (IBD). AI enables standardized, accurate, and timely disease assessment and outcome prediction, which are crucial for personalized treatment strategies. The integration of multi-OMICs data enhances patient profiling and management, demonstrating the potential of AI to refine risk stratification and improve therapeutic precision. This paradigm shift towards AI-enabled personalized interventions highlights the broader applicability of AI-driven personalized systems beyond education, into healthcare and other sectors.41
In conclusion, AI-driven personalized learning systems are at the forefront of transforming educational and clinical practices by providing tailored experiences that cater to individual needs. While the benefits are substantial, ongoing research and development are essential to overcome existing challenges and ensure ethical and effective implementation of these technologies across various domains.
In recent years, the application of large language models (LLMs)in medical education has attracted wide attention, especially in the field of physiology education. These models have brought profound changes to medical education through their powerful text processing and generation capabilities.
First, the application of Large Language Models (LLMs)in medical education can significantly enhance teaching quality and the design of personalized learning paths. By offering an interactive learning environment, LLMs can help students better grasp complex physiological concepts and foster critical thinking.42 Furthermore, the use of LLMs in medical education can optimize the teaching evaluation process, improve the efficiency of medical research, and support continuing medical education.12
However, the use of LLMs also presents several challenges. For example, issues with information accuracy, over-reliance on technology, a lack of emotional recognition, and concerns about ethics, privacy, and data security.43 To maximize the potential of LLMs and address these challenges, educators need to demonstrate leadership in medical education, adapt teaching strategies flexibly, foster students and 039; critical thinking, and emphasize the importance of practical experience.42
Furthermore, the application of LLMs in physiological education must also address potential biases and ethical concerns that may arise during content generation. To ensure the responsible and safe use of LLMs in medical education, it is recommended to develop a unified ethical framework tailored specifically for this field.43 This framework should be grounded in eight core principles: quality control and oversight mechanisms, privacy and data protection, transparency and explainability, fairness and equal treatment, academic integrity and ethical standards, accountability and traceability, protection and respect for intellectual property, and the promotion of educational research and innovation.43
In summary, the application of Large Language Models (LLMs)in medical education offers a promising opportunity to enhance the learning experience.13 However, ensuring the accuracy of information, emphasizing skill development, and maintaining ethical standards are essential. Continuous critical evaluation and interdisciplinary collaboration are crucial for the appropriate integration of LLMs into medical education.12 Through these efforts, LLMs can advance medical education while upholding principles of fairness, justice, and patient safety,43 thereby creating a more equitable, safer, and more efficient medical education environment.
Data privacy is a major concern in AI applications within medical physiology education. In the era of AI in healthcare, protecting patient safety and privacy is crucial.44 Issues such as patient privacy and confidentiality, protection of patient autonomy and informed consent, and the potential propagation of health care disparities need to be addressed.
In the context of ubiquitous health, a trust information-based privacy architecture has been proposed to help data subjects manage information privacy in an open and unsecure information space.45 This architecture enables the data subject to define service- and system-specific rules for data processing, providing reliable trust information and protection against privacy threats.
In China’s medical education, the application of artificial intelligence (AI) is becoming increasingly widespread. However, the data security challenges that come with it cannot be overlooked. The use of AI in medical education assessment poses a threat to academic integrity. AI chatbots may be misused for cheating, leading some institutions to implement internet restrictions.46 To address this issue, five strategies have been proposed: on-site proctoring, online proctoring, clear expectations and consequences, building an integrity-focused institutional culture, and reducing exam pressure. However, these strategies have their limitations, and there may be a need to consider formulating relevant regulations, fostering collaboration between medical educators and AI developers, and rethinking medical education more broadly.46
Moreover, the application of AI in medical data management faces challenges related to data integrity, accuracy, consistency, and precision. In the context of China’s medical information data management, a medical metadata governance framework has been proposed to achieve scientific governance of clinical data and ensure information privacy and security.47 This framework integrates metadata and master data management, aiming to guide the identification, cleaning, mining, and deep application of electronic medical record (EMR) data, thereby addressing the bottlenecks in current medical scenarios and supporting more effective clinical research and data-driven decision-making.47
To sum up, the application of AI in medical education and medical informatization in China has great potential, but it also brings challenges to data security and academic integrity. By formulating appropriate governance frameworks and strategies, we can make full use of the advantages of AI technology while ensuring data security and academic integrity.
Balancing traditional and AI-based teaching methods is essential for effective medical physiology education. A review of innovative teaching methods in radiology education highlighted the need to add new methods, such as audience response technology, long-distance teaching, the flipped classroom, and active learning, to the traditional lecture-type teaching format.48
In a comparison of traditional and interactive teaching methods in an emergency department, while both methods showed no significant difference in post-tutorial knowledge, students in the interactive multimedia-based tutorial group had a slightly lower satisfaction rate (89% vs 100% in the lecture group).49 This indicates that while AI-based interactive methods have potential, traditional methods still hold value, and a balance between the two is needed.
The challenges and controversies in AI integration in medical physiology education, including ethical considerations, data privacy concerns, and the need to balance traditional and AI-based teaching methods, must be addressed. Ethical frameworks need to be developed and implemented, data privacy protection mechanisms must be strengthened, and a thoughtful approach to combining traditional and new teaching methods is required to ensure the sustainable and effective integration of AI in medical education.
China’s’ master-apprentice teaching’ is a unique educational model, especially in the field of traditional Chinese medicine (TCM). This model passes on rich clinical experience and knowledge to the next generation through a master-apprentice tradition. However, with the rapid advancement of artificial intelligence (AI) technology, how to effectively integrate AI tools into this traditional teaching model has become a topic worthy of in-depth discussion.
In some traditional Chinese medicine hospitals, efforts are underway to transform the rich clinical experience of senior professors into an AI knowledge base. This process not only digitizes traditional practices but also innovates teaching methods. By leveraging AI technology, young teachers can more intuitively understand the diagnostic and decision-making processes of their senior colleagues, thereby enhancing both teaching efficiency and quality. This integration not only helps preserve and pass on the valuable experience of traditional Chinese medicine but also aids young teachers in better understanding and applying this knowledge with AI support.50
Furthermore, the application of AI in traditional Chinese medicine (TCM) education can leverage machine learning and deep learning technologies to analyze vast amounts of clinical data. This aids teachers and students in better understanding the complex theories and diagnostic methods of TCM. For instance, AI-assisted tongue diagnosis, pulse diagnosis, and syndrome identification can provide more objective and standardized references for teaching, thereby enhancing the scientific rigor and accuracy of the educational process.50
In short, integrating AI tools into the unique “mastering system teaching” in China is not only a reform of the traditional teaching mode, but also an important attempt to modernize TCM education. This integration is expected to bring more possibilities and innovation space for TCM teaching in the future.
The future of AI in medical physiology education holds great potential for innovation. In the field of neurosurgery, AI has the potential to be used as a diagnostic, predictive, intraoperative, or educational tool.51 For example, AI-powered systems can provide real-time feedback to surgeons, enhancing precision and reducing the risk of complications during surgical procedures.
In endodontic education, AI chatbots can offer personalized learning, interactive training, and clinical decision support.36 However, challenges such as technical limitations, ethical considerations, and the potential for misinformation need to be addressed. Despite these challenges, the development of such AI-based educational tools can revolutionize the way medical students learn complex physiological and clinical concepts.
The long-term impacts of AI on medical training and healthcare delivery are expected to be significant. In medical training, AI can potentially shape future health professions students by providing them with new learning opportunities and experiences.52 For example, a medical scribe fellowship program was found to be feasible and cost-effective, and it could potentially help students gain skills and better position themselves for professional schooling.
In healthcare delivery, immersive technologies such as VR and AR are expected to enhance patient care and medical training.53 These technologies can provide immersive and interactive environments for learning and practice, improving the skills and knowledge of medical professionals and ultimately leading to better patient outcomes.
The future prospects of AI in China’s medical physiology education are promising, with potential innovations in educational tools, the development of policy and regulatory frameworks, and significant long-term impacts on medical training and healthcare delivery. However, to fully realize these prospects, continuous efforts are needed to address the associated challenges, ensure ethical and regulatory compliance, and promote the effective integration of AI in the medical education and healthcare ecosystem.
In China, with the rapid advancement of generative artificial intelligence (GAI), medical education is facing unprecedented opportunities and challenges. To effectively integrate GAI into medical education, it is crucial to develop relevant policies and practical approaches. First, medical educational institutions must establish appropriate policies to ensure that the use of GAI complies with ethical standards and protects the data security of both institutions and patients. These policies should not only provide clear guidance on the proper use of AI in education but also encourage students to participate as co-inventors of local innovations.54
Secondly, establishing a GAI governance mechanism is essential for guiding the ethical and fair use of GAI. Through this mechanism, educational institutions can establish guidelines to ensure that the use of GAI is not only technologically advanced but also socially responsible. Additionally, to ensure the safety of GAI experiments for teachers and students, it is essential to provide the necessary tools and training. This will help define the skill requirements for students and teachers in GAI and ensure their competitiveness in the rapidly evolving medical landscape.54
Finally, the design of GAI-related courses should emphasize their impact in clinical settings. Educational institutions should continuously update the expected competencies of medical students to align with the latest advancements in GAI education and prepare them for a career that is continually evolving due to GAI. By sharing best practices and collecting data to assess the impact of GAI education, the medical education community can collaborate to ensure the effective integration of GAI into medical education.54
AI is playing an increasingly important role in understanding disease mechanisms in physiology. By integrating proteomics of liquid biopsies with single-cell transcriptomics, researchers have been able to trace the cellular origin of proteins in the aqueous humor, leading to new insights into eye aging and disease mechanisms.55 This approach can potentially transform molecular diagnostics and prognostics in ophthalmology.
In the study of gene regulatory networks, a novel algorithm called Dandelion has been developed to construct and train interspecies Bayesian networks.56 This algorithm can identify key components of disease networks, providing crucial information on regulatory relationships among genes and leading to a better understanding of disease molecular mechanisms, such as in oculopharyngeal muscular dystrophy.
In China, AI is making significant contributions to epidemiological studies. For example, in the study of brucellosis, the isolation and identification of a novel Brucella abortus biovar 4 strain from yak in Tibet was reported.57 This finding, which was aided by various molecular techniques, can help improve the diagnosis and epidemiological understanding of brucellosis in the region.
In the analysis of spatial delays of tuberculosis in Eastern China, spatial-temporal scan statistics and other methods were used to identify clusters of long-term patient and diagnostic delays.58 This type of analysis, which could potentially be enhanced with AI-based data analytics, can help in formulating targeted public health strategies to address tuberculosis in the region.
AI is enhancing diagnostic techniques in physiology education. In the field of radiology, AI-assisted CT diagnostic technology for the classification of pulmonary nodules has shown good diagnostic performance, with a pooled sensitivity of 0.90 and specificity of 0.89.59 However, its specificity needs to be further improved.
In medical education, the integration of AI into biomedical science curricula is crucial to prepare future healthcare workers.60 For example, teaching students about AI-related informatics, data sciences, and digital health can enhance their understanding of how AI can be used to improve diagnostic techniques in physiology and other medical fields.
AI is providing valuable insights into disease mechanisms, contributing to epidemiological studies in China, and enhancing diagnostic techniques in physiology education. However, there is still room for improvement, such as improving the specificity of diagnostic technologies and further integrating AI into medical curricula. Continued research and development in these areas will be essential for leveraging AI to improve medical understanding, diagnosis, and ultimately patient care.
This review aimed to explore the integration of artificial intelligence (AI) into China’s medical physiology education, examining its pedagogical practices, systemic challenges, and transformative potential. Key findings reveal that AI-driven tools—including machine learning for physiological simulations, immersive VR/AR platforms, and personalized learning systems—significantly enhance diagnostic accuracy, student engagement, and adaptive training outcomes. However, challenges such as algorithmic bias, data privacy risks, and resource disparities between urban and rural institutions, alongside tensions between AI efficiency and humanistic pedagogy, remain critical barriers to widespread adoption.
These insights underscore the need for a balanced approach to AI integration: one that leverages technological innovation while preserving core educational values. Future efforts should prioritize the development of robust ethical frameworks to ensure transparency and equity, equitable resource distribution to bridge regional gaps, and interdisciplinary collaboration among educators, policymakers, and technologists. By harmonizing AI capabilities with pedagogical integrity, Chinese medical physiology education can cultivate a new generation of clinicians equipped with both technical proficiency and ethical discernment, ultimately advancing healthcare quality and accessibility (Table 1).
|
Table 1 Framework Synthesis of AI Integration in China’s Medical Physiology Education
|
This work was supported by the basic research project of Education Department of Yunnan Province.
This study was supported by the Science and Education Project of Yunnan Province (No. 2024J0064), and Joint Medical Program of KMUST and Affiliated Hospital (KUST-AN2023002Q).
The author reports no conflicts of interest in this work.
1. Silver JK, Dodurgali MR, Gavini N. Artificial intelligence in medical education and mentoring in rehabilitation medicine. AmerJ Phys Med Rehab. 2024;103(11):1039–1044. doi:10.1097/PHM.0000000000002604
2. Anderson TR, Rogan JM. Bridging the educational research-teaching practice gap curriculum development, part 1: components of the curriculum and influences on the process of curriculum design. Biochem Mol Biol Educ. 2011;39(1):68–76.
3. Gahutu JB. Physiology teaching and learning experience in a new modular curriculum at the national university of Rwanda. Adv Physiolog Educ. 2010;34(1):11–14. doi:10.1152/advan.00093.2009
4. Qiu TT, et al. International curriculum for undergraduate sonographer education in China during the COVID-19 era: international remote teaching mode vs. domestic on-site teaching mode. Front Public Health. 2022;10:1083108.
5. Rodrigues MAM, Silva RR, Santos DAT, et al. Investigating academic performance and perceptions of human physiology and exercise physiology courses among undergraduate students of physical education at a Brazilian public university. Bmc Med Educ. 2024;24(1). doi:10.1186/s12909-024-05058-1.
6. D’Souza RF, Mathew M, Mishra V, et al. Twelve tips for addressing ethical concerns in the implementation of artificial intelligence in medical education. Med Educ Online. 2024;29(1).
7. Naseer MA, Saeed S, Afzal A, et al. Navigating the integration of artificial intelligence in the medical education curriculum: a mixed-methods study exploring the perspectives of medical students and faculty in Pakistan. Bmc Med Educ. 2025;25(1). doi:10.1186/s12909-024-06552-2.
8. Li RL, Liu G, Zhang M, et al. Integration of artificial intelligence in radiology education: a requirements survey and recommendations from faculty radiologists, residents, and medical students. Bmc Med Educ. 2025;25(1). doi:10.1186/s12909-025-06859-8.
9. Ramos CJ, Cartagena K, Ramirez KA, et al. Analyzing the role of ai in medical education: an evaluation on the performance of chatgpt on ophthalmology trainee examination questions. Invest Ophthalmol Visual Sci. 2024;65(7):359.
10. Su WH, Zhang ST, Chen SH. Prospects for the development of medical professionalism education in the AI perspective: a qualitative study of Chinese postgraduate medical students’ written reflections. Med Teach. 2025:1–9.
11. Gin BC, O’Sullivan PS, Hauer KE, et al. Entrustment and EPAs for artificial intelligence (AI): a framework to safeguard the use of AI in health professions education. Acad Med. 2025;100(3):264–272. doi:10.1097/ACM.0000000000005930
12. Lucas HC, Upperman JS, Robinson JR. A systematic review of large language models and their implications in medical education. Med Educ. 2024;58(11):1276–1285. doi:10.1111/medu.15402
13. Kiyak YS, Emekli E. ChatGPT prompts for generating multiple-choice questions in medical education and evidence on their validity: a literature review. Postgraduat Med J. 2024;100(1189):858–865. doi:10.1093/postmj/qgae065
14. Singapogu R, Chowdhury A, Roy-Chaudhury P, et al. Simulator-based hemodialysis cannulation skills training: a new horizon? Clin Kidney J. 2021;14(2):465–470. doi:10.1093/ckj/sfaa206
15. Aurello P, Pace M, Goglia M, et al. Enhancing Surgical Education Through Artificial Intelligence in the Era of Digital Surgery. American Surgeon. 2025. doi:10.1177/00031348251346539
16. Bahari G, Alenazi L, Alkathiri AM. The implementation of artificial intelligence in nursing education and its psychological impact on nursing students: a cross-sectional study. Nurs Educ Today. 2025;153.
17. Liu H, Mi XF, Huang ZZ, et al. Challenges and strategies in developing team-based learning in Chinese medical education. Medical Teacher. 2020;42(11):1243–1249. doi:10.1080/0142159X.2020.1801995
18. Zhang LQ, Xiao M, Gu M, et al. An overview of the roles and responsibilities of chinese medical colleges in body donation programs. Anatom Sci Educ. 2014;7(4):312–320. doi:10.1002/ase.1422
19. Porto JR, Morgan KA, Hecht CJ, et al. Quantifying the scope of artificial intelligence-assisted writing in orthopaedic medical literature: an analysis of prevalence and validation of AI-detection software. J Amer Acad Orthop Surg. 2025;33(1):42–50. doi:10.5435/JAAOS-D-24-00084
20. Li L, Guo J, Wu X, et al. Does promoting new urbanization help mitigate environmental pollution: evidence from China’s pilot policies for new urbanization. J Environ Manage. 2025;388:125925.
21. Busch F, Hoffmann L, Truhn D, et al. Global cross-sectional student survey on AI in medical, dental, and veterinary education and practice at 192 faculties. Bmc Med Educ. 2024;24(1). doi:10.1186/s12909-024-06035-4.
22. Hughes JDM, Azzi E, Rose GW, et al. A survey of senior medical students’ attitudes and awareness toward teaching and participation in a formal clinical teaching elective: a Canadian perspective. Med Educ Online. 2017;22(1):1270022.
23. Duan S, Liu C, Rong T, et al. Integrating AI in medical education: a comprehensive study of medical students’ attitudes, concerns, and behavioral intentions. Bmc Med Educ. 2025;25(1). doi:10.1186/s12909-025-07177-9.
24. Chen MY, Zhang B, Cai Z, et al. Acceptance of clinical artificial intelligence among physicians and medical students: a systematic review with cross-sectional survey. Front Med. 2022;9:990604.
25. Dos Santos DP, Giese D, Brodehl S, et al. Medical students’ attitude towards artificial intelligence: a multicentre survey. Eur Radiol. 2019;29(4):1640–1646. doi:10.1007/s00330-018-5601-1
26. Huang WN, Li Y, Bao Z, et al. Knowledge, attitude and practice of radiologists regarding artificial intelligence in medical imaging. J Multidisciplin Healthcare. 2024;17:3109–3119. doi:10.2147/JMDH.S451301
27. Collado-Mesa F, Alvarez E, Arheart K. The role of artificial intelligence in diagnostic radiology: a survey at a single radiology residency training program. J Am College Radiol. 2018;15(12):1753–1757. doi:10.1016/j.jacr.2017.12.021
28. Jammal AA, Thompson AC, Mariottoni EB, et al. Human versus machine: comparing a deep learning algorithm to human gradings for detecting glaucoma on fundus photographs. Am J Ophthalmol. 2020;211:123–131. doi:10.1016/j.ajo.2019.11.006
29. Alexiadis A. Deep multiphysics: coupling discrete multiphysics with machine learning to attain self-learning in-silico models replicating human physiology. Artif Intell Med. 2019;98:27–34. doi:10.1016/j.artmed.2019.06.005
30. Song HX, Sun Z, Hao Y, et al. Machine learning-assisted prediction and optimization of electrocatalytic activity. J Mater Sci Technol. 2026;245:17–32. doi:10.1016/j.jmst.2025.04.050
31. Zou LX, Wang X, Hou Z-L, et al. Machine learning algorithms for diabetic kidney disease risk predictive model of Chinese patients with type 2 diabetes mellitus. Renal Failure. 2025;47(1). doi:10.1080/0886022X.2025.2486558.
32. Tsang KCH, Pinnock H, Wilson AM, et al. Application of machine learning algorithms for asthma management with mHealth: a clinical review. J Asthma Allergy. 2022;15:855–73.
33. Goo HW, Park SJ, Yoo SJ. Advanced medical use of three-dimensional imaging in congenital. heart disease: augmented reality, mixed reality, virtual reality, and three-dimensional printing. Korean J Radiol. 2020;21(2):133–145. doi:10.3348/kjr.2019.0625
34. Moro C, Birt J, Stromberga Z, et al. Virtual and augmented reality enhancements to medical and science student physiology and anatomy test performance: a systematic review and meta-analysis. Anatom Sci Educ. 2021;14(3):368–376. doi:10.1002/ase.2049
35. Dhar P, Rocks T, Samarasinghe RM, et al. Augmented reality in medical education: students’ experiences and learning outcomes. Medical Education Online. 2021;26(1). doi:10.1080/10872981.2021.1953953.
36. Mondal H, Mondal S. Adopting augmented reality and virtual reality in medical education in resource-limited settings: constraints and the way forward. Adv Physiolog Educ. 2025;49(2):503–507. doi:10.1152/advan.00027.2025
37. Zayet TMA, Ismail MA, Almadi SH, et al. What is needed to build a personalized recommender system for K-12 students’ E-learning? Recommendations for future systems and a conceptual framework. Educat Inform Technol. 2022;28(6):7487–508.
38. Ahmed ST, Mahesh TR, Srividhya E, et al. Towards blockchain based federated learning in categorizing healthcare monitoring devices on artificial intelligence of medical things investigative framework. BMC Med Imaging. 2024;24(1). doi:10.1186/s12880-024-01279-4.
39. Komolafe OO, Mustofa J, Daley MJ, et al. Current applications and outcomes of AI-driven adaptive learning systems in physical rehabilitation science education: a scoping review protocol. PLoS One. 2025;20(6):e0325649. doi:10.1371/journal.pone.0325649
40. Xu XW. AI optimization algorithms enhance higher education management and personalized teaching through empirical analysis. Sci Rep. 2025;15(1):10157.
41. Li A, Wang YX, Chen HX. AI driven cardiovascular risk prediction using NLP and large language models for personalized medicine in athletes. Slas Technol. 2025;32:100286.
42. Koçak M, Oguz AK, Akçali Z. The role of artificial intelligence in medical education: an evaluation of large language models (LLMs) on the Turkish medical specialty training entrance exam. Bmc Med Educ. 2025;25(1). doi:10.1186/s12909-025-07148-0
43. Li Z, Fenghe L, Xuehu W, et al. Ethical considerations and fundamental principles of large language models in medical education: viewpoint. J Med Internet Res. 2024;26:e60083.
44. Alonso A, Siracuse JJ. Protecting patient safety and privacy in the era of artificial intelligence. Semin Vasc Surg. 2023;36(3):426–429. doi:10.1053/j.semvascsurg.2023.06.002
45. Ruotsalainen PS, Blobel B, Seppälä A, et al. Trust information-based privacy architecture for ubiquitous health. Jmir mHealth and uHealth. 2013;1(2):e23. doi:10.2196/mhealth.2731
46. Sridharan K, Sequeira RP. Artificial intelligence and medical education: application in classroom instruction and student assessment using a pharmacology & therapeutics case study. Bmc Medical Education. 2024;24(1). doi:10.1186/s12909-024-05365-7
47. Lu L, Zhong Y, Luo S, et al. Dilemmas and prospects of artificial intelligence technology in the data management of medical informatization in China: a new perspective on SPRAY-type AI applications. Health Informatics J. 2024;30(2). doi:10.1177/14604582241262961.
48. Sivarajah RT, Curci NE, Johnson EM, et al. A review of innovative teaching methods. Acad Radiol. 2019;26(1):101–113. doi:10.1016/j.acra.2018.03.025
49. Armstrong P, Elliott T, Ronald J, et al. Comparison of traditional and interactive teaching methods in a UK emergency department. Eur J Emerg Med. 2009;16(6):327–329. doi:10.1097/MEJ.0b013e32832b6375
50. Pan DP, Guo Y, Fan Y, et al. Development and application of traditional chinese medicine using ai machine learning and deep learning strategies. Ame J Chin Med. 2024;52(03):605–623. doi:10.1142/S0192415X24500265
51. Windermere SA, Shah S, Hey G, et al. Applications of artificial intelligence in neurosurgery for improving outcomes through diagnostics, predictive tools, and resident education. World Neurosurg. 2025;197.
52. Lin S, Duong A, Nguyen C, et al. Five years’ experience with a medical scribe fellowship: shaping future health professions students while addressing provider burnout. Acad Med. 2021;96(5):671–679. doi:10.1097/ACM.0000000000003757
53. Khan HU, Ali Y, Khan F, et al. A comprehensive study on unraveling the advances of immersive technologies (VR/AR/MR/XR) in the healthcare sector during the COVID-19. Challeng Solut Heliyon. 2024;10(15).
54. Triola MM, Rodman A. Integrating generative artificial intelligence into medical education: curriculum, policy, and governance strategies. Acad Med. 2025;100(4):413–418. doi:10.1097/ACM.0000000000005963
55. Wolf J, Rasmussen DK, Sun YJ, et al. Liquid-biopsy proteomics combined with AI identifies cellular drivers of eye aging and disease. Cell. 2023;186(22):4868. doi:10.1016/j.cell.2023.09.012
56. Anvar SY, Tucker A, Vinciotti V, et al. Interspecies translation of disease networks increases robustness and predictive accuracy. PLOS Computatl Biolog. 2011;7(11). doi:10.1371/annotation/fc0b4192-6427-4fb3-b347-c66651adf855.
57. Sun MJ, Liu M, Zhang X, et al. First identification of a biovar 4 strain from yak in Tibet, China. Vet Microbiol. 2020;247:108751.
58. Luo D, Chen X, Wang M, et al. Analyzing spatial delays of tuberculosis from surveillance and awareness surveys in Eastern China. Sci Rep. 2024;14(1).
59. Huang G, Wei X, Tang H, et al. A systematic review and meta-analysis of diagnostic performance and physicians’ perceptions of AI-assisted CT diagnostic technology for the classification of pulmonary nodules. J Thoracic Dis. 2021;13(8):4797. doi:10.21037/jtd-21-810
60. Oliveira PBD, Cunha JB. Blending artificial intelligence into PID controller design: a. Biomed Enginee Exper Ifac Papersonline. 2016;49(6):366–371. doi:10.1016/j.ifacol.2016.07.205
The long-awaited Draw No. 103 for the Rs 1,500 prize bond was held today at the National Savings Centre in Faisalabad. The draw, which began at 10 AM, delivered some exciting news for lucky participants nationwide.
The first prize of Rs 3,000,000 was followed by three second prizes worth Rs 1,000,000 each. 1,696 holders secured the Third Prize of Rs 18,500 each. The full list of third-prize winners is available on official National Savings platforms.
Prize Bond: Rs 1,500
Draw Number: 103
Draw Date: 15-08-2025
First Prize – Rs 3,000,000 (One Winner):
790468
Second Prize – Rs 1,000,000 (Three Winners):
607650
193673
031085
To claim, winners must present the original bond along with their Computerized National Identity Card (CNIC). The Rs 1,500 prize bond remains one of the most popular investment tools in Pakistan, valued for its low cost and high payout potential. Held quarterly, these draws offer everyday citizens a shot at life-changing sums, with strong participation from every province.
Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection, and it remains a clinical syndrome with a high mortality rate.1 Approximately 20–30% of patients in intensive care units develop sepsis.2 According to data from the Global Burden of Disease study in 2017, there were 48.9 million cases of sepsis worldwide.3 Although the age-standardized incidence and mortality rates have declined, sepsis continues to be a leading cause of global health loss, with the highest burden observed in sub-Saharan Africa, Oceania, South Asia, East Asia, and Southeast Asia.2,3 Current clinical criteria for sepsis diagnosis integrate the SOFA (Sequential Organ Failure Assessment) score with laboratory biomarkers. While the SOFA score is employed to quantify organ dysfunction and predict mortality, it suffers from delayed diagnosis requiring serial monitoring. In contrast, the qSOFA (quick SOFA) score enables rapid bedside assessment of disease severity but demonstrates reduced sensitivity and unreliability in immunocompromised patients.4 Supplementary biomarkers such as procalcitonin (PCT), C-reactive protein (CRP), lactate, and blood cultures exhibit inherent limitations. These include insufficient timeliness (eg, delayed elevation of PCT/CRP leading to treatment delays), limited specificity, and an inability to differentiate immune endotypes. Collectively, these issues elevate the risk of misdiagnosis. Recent studies further highlight the urgent need to develop novel sepsis biomarkers to enhance capabilities for early diagnosis and precision therapeutics.5 Novel biomarkers aim to address these unmet clinical needs by potentially offering enhanced specificity, facilitating the differentiation of sepsis from non-infectious syndromes, enabling early detection via rapid pathophysiological changes, and allowing for immune phenotypic stratification. This capability ensures more timely interventions and improved clinical outcomes. Consequently, this study focuses on characterizing such biomarkers to overcome the limitations inherent in current diagnostic approaches. At present, the treatment of sepsis mainly involves three aspects: infection control, hemodynamic management, and modulation of the host response. Infection control applies to all sepsis cases.6 Due to the heterogeneity of the sepsis process, the diagnosis and definition of sepsis remain key issues in clinical practice,6,7 therefore, underscoring the urgent need for improvements in sepsis diagnosis and treatment.
Parthanatos, a form of programmed cell death distinct from apoptosis, necrosis, and autophagy, is triggered by DNA damage and mediated by poly ADP-ribose polymerase 1 (PARP1).8 The main mechanisms include DNA damage, overactivation of PARP1, accumulation of PAR, depletion of nicotinamide adenine dinucleotide (NAD) and adenosine triphosphate (ATP), and nuclear translocation of apoptosis inducing factor (AIF).9 Parthanatos has been considered a promising therapeutic target in cancer.10 It has been reported in various diseases, including ischemic stroke and neurodegenerative diseases.11 In cancer, parthanatos has been linked to diseases such as hepatocellular carcinoma and triple-negative breast cancer.10 In the context of sepsis, Lorente L. et al investigated the relationship between parthanatos and sepsis mortality, finding an association between parthanatos and sepsis-related deaths,12 although the underlying molecular mechanisms remain unclear. Furthermore, it has been shown that caspase-11 parthanatos leads to pathological changes and shortens survival time in sepsis models. However, parthanatos has been less studied in the development of sepsis and further research is needed.
Mendelian Randomization (MR) is a novel epidemiological approach that uses single nucleotide polymorphisms (SNPs) as instrumental variables to infer causal relationships between exposures and outcomes based on genome-wide sequencing data.13 MR is built upon three key assumptions: (1) the genetic variant is strongly associated with the exposure (relevance assumption), (2) the genetic variant is not associated with confounders of the exposure-outcome relationship (independence assumption), and (3) the genetic variant influences the outcome solely through the exposure (exclusivity assumption).14 Compared to randomized controlled trials, MR offers statistical advantages by mitigating bias and reverse causation, while being more cost-effective.15
In summary, this study utilized sepsis transcriptome data from the Gene Expression Omnibus (GEO) database and 11 parthanatos-related genes (PRGs) from the literature to obtain biomarkers related to parthanatos in sepsis by differential expression analysis, weighted gene co-expression network analysis (WGCNA), MR analysis and machine learning. The diagnostic potential and underlying molecular mechanisms of these biomarkers were further analyzed, providing new insights into the clinical diagnosis, prevention, and treatment of sepsis.
GSE65682 (GPL13667 platform) served as the training cohort, and GSE95233 (GPL570 platform) was used as the validation cohort, both of which were extracted from the GEO database (http://www.ncbi.nlm.nih.gov/geo/). GSE65682 comprised whole blood samples from 760 individuals diagnosed with sepsis and 42 samples from healthy individuals. Additionally, GSE95233 consisted of whole blood samples from 102 patients with sepsis alongside 22 samples from control subjects. The GSE167363 dataset was extracted from the GEO database for single-cell analysis, including 6 human peripheral blood mononuclear cell (PBMC) samples, comprising 2 healthy controls, 2 sepsis survivors, and 2 sepsis non-survivors. The parthanatos-related genes (PRGs) were collected from relevant literature.16 The sepsis dataset (ieu-b-4980) was gotten from Integrative Epidemiology Unit (IEU) Open Genome-Wide Association Study (GWAS) database (https://gwas.mrcieu.ac.uk/), including 486,484 samples of Europeans (11,643 sepsis and 474,841 control samples) and 12,243,539 SNPs. Acquisition of expression quantitative trait loci (eQTL) data of candidate genes as exposure factors via GWAS database.
Differential expression analysis in training set was performed utilizing limma package (v 3.50.1)17 to derive differentially expressed genes (DEGs) between sepsis and control samples (padj < 0.05 and |log2Fold Change (FC)| > 0.5). Subsequently, DEGs were visualized by ggplot2 (v 3.4.4)18 pheatmap (v 1.0.12).19 The top 10 genes were annotated in each of the upregulated and downregulated DEGs, and these twenty genes were selected for heatmap.
WGCNA package (v 1.7.1)20 was used to analyze gene modules associated with PRGs. Initially, we calculated the scores for PRGs in both sepsis patients and healthy samples via single-sample gene set enrichment analysis (ssGSEA) algorithm. Then, the GoodSamplesGenes function was utilized to cluster and exclude outlier samples. PckSoftThreshold function was employed to determine the soft value required to construct the scale-free network (R2 = 0.8620). Following this, the Dynamic Tree Cutting method was implemented to construct the hierarchical clustering tree. Finally, we established a correlation analysis by calculating the matrix of correlation coefficients between the eigenvectors of the modules and the ssGSEA scores.
Candidate genes were pinpointed by intersecting DEGs with the genes from the key modules by the ggVenn package (v 0.1.9).21 Subsequently, in order to understand the possible pathways and biological functions of these candidate genes, we conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on candidate genes through clusterProfiler (v 4.7.1.003)22 (p < 0.05). Then, we investigated the interactions among the candidate genes on proteomic level, the genes were inputted into the STRING database for the acquisition of protein-protein interaction (PPI) data. (confidence > 0.4), and the network of candidate genes was then visualised using Cytoscape software (v 3.10.1).23
To identify candidate genes that had a causal relationship with sepsis, the TwoSampleMR package (v 0.6.4)24 extract_instruments function was harnessed for reading exposure factors and identify SNPs that were independently associated with these factors (p < 5×10-8, R²< 0.001, distance > 100kb and removed SNPs with F-statistics < 10). These SNPs were later applied as instrumental variables (IV) to determine the causality between the exposure and the resulting outcome. Subsequently, harmonise_data function was applied to standardize the effect alleles and effect sizes, and MR analysis was carried out combining 5 algorithms [MR Egger,25 Weighted median,26 Inverse variance weighted (IVW),27 Weighted mode,28 Simple mode].29 Among these, the IVW algorithm requires attention (p<0.05, OR ≠ 1) when assessing the results of the five algorithms. Following this, scatter plots (the negative slope suggested that the gene acts as a protective gene against Sepsis, while the positive slope suggested it as a risk factor for the condition), forest plots (the genes with effect size to the left of the dashed line are considered protective genes, whereas those to the right are identified as risk genes), and funnel plots (If the samples were evenly distributed around the IVW line,it would imply that MR adheres to Mendel’s second law) were utilized to further illustrate the MR results.
Heterogeneity test was conducted via mr_heterogeneity function. (Q_pval > 0.05), mr_pleiotropy_test function and the mr_presso function from the MRPRESSO package (v 1.0)30 were used to perform horizontal pleiotropy tests (p >0.05), and the mr_leaveoneout function performs leave-one-out analysis, which involves sequentially removing each SNP to see if the results change. Finally, the Steiger directionality test was employed to verify that the positive analysis structure was not affected by reverse causal effects, which was done by directionality_test function (steiger pval < 0.05). Genes analyzed by MR were defined as candidate signature genes for further research.
After identifying the candidate signature genes, the least absolute shrinkage and selection operator (LASSO) and Boruta algorithms were instrumental into filter out candidate biomarkers. The intersection of the results from both methods was used to determine the candidate biomarkers. The LASSO was completed through glmnet package (v 4.1–4),31 and the Boruta analysis was accomplished with the Boruta package (v 8.0.0).32 Following this, within GSE65682 and GSE95233 cohorts, performed expression profiling and receiver operating characteristic (ROC) analysis on the candidate biomarkers, ROC curves were generated through pROC package (v 1.18.0).33 Candidate biomarkers with consistent expression trends, significant differences in expression profiles, and the area under the curve (AUC) greater than 0.7 were selected as biomarkers for subsequent analysis.
To assess the predictive value of the biomarkers in Sepsis, a nomogram based on the biomarkers was constructed using the rms package (v 6.5–0).34 Subsequently, the accuracy of the nomogram’s predictive ability was evaluated through calibration plots and decision curve analysis. The accuracy of the nomogram was further verified using the ROC. The ggDCA package (v 1.2)35 was used for plotting the decision curves.
To identify the signaling pathways in which the biomarkers are involved, based on the sepsis samples and healthy samples in GSE65682, the Spearman correlation was calculated between the biomarkers and all other genes. The genes were then ranked according to the correlation coefficients. Subsequently, the clusterProfiler (v 4.7.1.003) was used to perform GSEA, and the reference gene set chosen was “c2.cp.kegg.v2023.1.Hs.symbols.gmt” (p < 0.05 and false discovery rate (FDR) < 0.25). Selected the top 5 enriched pathways from the GSEA results for presentation.
Subsequently, we investigated the differences in immune cell infiltration in GSE65682 by utilizing the CIBERSORT algorithm to quantify the levels of 22 specific immune cell types, then applied to evaluate disparities in the expression of different immune cell (p < 0.05). Thereafter, we explored the relationships between biomarkers, immune cells, and among immune cells themselves, conducting a Spearman correlation analysis on biomarkers and immune cells to assess their interrelatedness. (p < 0.05, |cor| > 0.3).
Molecular regulatory networks are key elements in the process of gene expression regulation, and they are of great research value as they can uncover the intricacies and variety of gene expression control mechanisms. Utilizing the biomarkers obtained, we leveraged the miRNet database (https://www.mirnet.ca/miRNet/docs/RTutorial.xhtml) to identify microRNAs (miRNAs) and long non-coding RNAs (lncRNAs) that were involved in the regulation of these biomarkers. We then filtered for miRNAs that simultaneously target the biomarkers and predicted the related lncRNAs using these miRNAs, using these as the core to build a lncRNA-miRNA-mRNA regulatory network. In order to examine the interplay between biomarkers and transcription factors (TFs), we utilized the miRNet database to forecast TFs that have interactive relationships with the biomarkers and established a miRNA-mRNA-TF regulatory network. Both of regulatory networks were completed using Cytoscape (v 3.10.1).
Subsequently, we conducted research on the biological characteristics of these biomarkers, utilizing GeneMANIA to construct and analyze the interaction network among the biomarkers as well as their co-expressed genes.
Gene localization is essential for exploring the structure, function, and interactions of genes, and it helps in understanding the impact of genetic factors on gene expression more profoundly. In order to understand the chromosomal locations of biomarkers, we employed the OmicCircos (v 1.36.0)36 to visualize the distribution of these biomarkers across the chromosomes and illustrate the expression profiles of biomarkers on various chromosomes.
Based on drug signature database (DsigDB) (http://dsigdb.tanlab.org/DSigDBv 1.0/), we utilized the enrichR package (v 3.1)37 to identify candidate drugs that may target the biomarkers (p < 0.05). Furthermore, we employed Cytoscape software to present a visualized network of interconnections between the biomarkers and a range of predicted drugs.
The single-cell transcriptomic sequencing data (GSE167363) were constructed into Seurat objects using the R package “Seurat” (v 5.1.0),38 with cells having fewer than 300 genes and genes covered by fewer than three cells excluded. The parameters were set to 6000 > nFeature_RNA > 200 and nCount_RNA < 25000. The data were then normalized. The FindVariableFeatures function was used to identify highly variable genes based on the relationship between the mean and variance of gene expression, with the top 2,000 highly variable genes selected for further analysis by default. The results were visualized using the LabelPoints function, with the top 10 most variable genes labeled. To further confirm and analyze the cell populations of different cell groups, the R package “Seurat” (v 5.1.0) was used to normalize the single-cell transcriptome sequencing data (GSE167363) using the ScaleData function. Principal component analysis (PCA) was then performed on the highly variable genes of each sample for dimensionality reduction. The Jackstraw function was used to generate a Jackstraw plot, and re-clustering was performed using a permutation test algorithm to select the appropriate principal components. The ElbowPlot function was used to create an elbow plot to identify the usable dimensions. Based on the plateau in the PCA feature number shown in the elbow plot and the analysis from the Jackstraw plot, suitable principal components were selected for further analysis. The FindNeighbors and FindClusters functions in Seurat (v 5.1.0) were used to identify different cell clusters, and uniform manifold approximation and projection (UMAP) was employed for visualization (resolution = 0.1). For cell type annotation within the clusters, the CellMarker2.0 database (http://bio-bigdata.hrbmu.edu.cn/CellMarker) and manual labeling were used to annotate each cell cluster. Cell types were annotated based on the marker genes provided in the literature.39 Finally, UMAP plots of the annotated cell clusters were generated, and a DotPlot was created to display the distribution of biomarkers across different groups and cell types. In addition, the expression of biomarkers in different cells was verified to identify key cells.
For cellular communication analysis, the CellChat package (v 1.6.1)40 was used, leveraging the CellChatDB database (https://www.cellchat.org/db/). This tool utilized cell expression data as input to model cell-cell communication through ligand-receptor and cofactor interactions. Additionally, pseudotime analysis was performed using the Monocle2 package (v 2.26.0)41 to explore key cellular transitions during different developmental stages, and DifferentialGeneTest was employed to analyze gene dynamics during the cellular differentiation process.
Analyzed the expression of biomarkers in blood samples from sepsis patients and control subjects through RT-qPCR. Single blinding was applied during the validation phase. Total RNA was extracted from blood samples of 5 sepsis patients and 5 control subjects using TRIzol reagent (Ambion, Austin, USA). After standing for 15 minutes, 1 μL was taken to quantify RNA concentration using a NanoPhotometer N50. Then, reverse transcription to synthesize cDNA was performed using the SureScript First strand cDNA synthesis kit (Servicebio, Wuhan, China). The obtained cDNA was diluted and then subjected to RT-qPCR experiments according to a certain system. The results were analyzed using the 2-ΔΔCt method, with GAPDH serving as the reference gene to ensure the accuracy of the experiment. The specific primer sequences, reaction systems, and amplification conditions are shown in Supplementary Tables 1–3.
All statistical analyses were conducted using R software (v 4.2.3). The Wilcoxon test was employed to assess differences between two groups, considering p < 0.05 as indicative of statistical significance.
Within the GSE65682, we identified 4,336 DEGs between sepsis patients and control groups, comprising 1,370 upregulated genes and 2,966 downregulated genes. The top ten genes from both the upregulated and downregulated DEGs were selected for marking and visualization through a heatmap (Figure 1A and B). Following that, we established a co-expression network for genes based on PRGs using WGCNA. At the outset, we analyzed the expression of PRGs in sepsis and control groups, selecting 9 PRGs with significant scoring differences (Figure 1C). A hierarchical clustering analysis was then conducted on all samples, from which we retained 801 samples (Figure 1D). We chose β= 8 to build a scale-free network (R² = 0.8620) (Figure 1E). After that, a hierarchical clustering tree was constructed to classify the genes and total of 17 co-expression modules were obtained (Figure 1F). Ultimately, the MEgreen module (cor = 0.63, p = 9e-91) as the key module for further analysis (Figure 1G). Finally, we retained 157 genes within the MEgreen module (Gene Significance (GS) > 0.2, Module Membership (MM) > 0.2) (Figure 1H).
|
Figure 1 Differential expression analysis. (A) Volcano map of differentially expressed genes; (B) Heat map of expression of differentially expressed genes; (C) Scoring violin plot for ssGSEA analysis. (D) Sample level clustering tree; (E) Selection of soft thresholds. (F) Identification of co-expression modules; (G) Heatmap of correlation between modular genes and single-sample Gene Set Enrichment Analysis (ssGSEA) score constructs; (H) Scatterplot of correlation between modules and traits.
|
In the Venn diagram, 84 candidate genes were acquired through intersection of DEGs and MEgreen module (Figure 2A). Following this, GO analysis showed a significant enrichment of these candidate genes across 119 distinct categories, including 9 Cellular Components (CC), 61 Molecular Functions (MF), and 49 Biological Processes (BP), CCs featured complexes like the serine/threonine protein kinase complex and the protein acetyltransferase complex, MFs involved binding activities such as vinculin binding and the activity of cysteine-type endopeptidases in the apoptotic process, BPs included the regulation of transcription elongation from a DNA template and the positive regulation of signaling pathways involving the transforming growth factor beta receptor (Figure 2B). Additionally, the candidate genes were found to be enriched in KEGG pathways such as the cell cycle, the FoxO signaling pathway, and the Apelin signaling pathway (Figure 2C). Next, to explore the interactions among the candidate genes, we established a PPI network that comprised 60 nodes and 72 edges., where we excluded 24 genes that did not have any interactions (Figure 2D).
 |
Figure 2 Acquisition and functional analysis of candidate genes. (A) Venn diagram; (B) Results of Gene ontology (GO) enrichment analysis; (C) Results of kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis. (D) Protein-Protein Interaction (PPI) network architecture.
|
To identify candidate genes that have a causal relationship with sepsis, we conducted MR analysis. In the IVW algorithm, 8 candidate genes were significantly causally associated with sepsis, where BRD1 (β= 0.127, p = 0.004, 95% confidence interval (CI) = 0.040–0.214), CASP2 (β= 0.217, p = 0.000, 95% CI = 0.110–0.323), HNRNPF (β= 0.067, p = 0.044, 95% CI = 0.002–0.132), and GCC1 (β= 0.111, p = 0.000, 95% CI = 0.057–0.166) were risk factors, while BTN2A1 (β= −0.037, p = 0.003, 95% CI = −0.061 to −0.013), MSL2 (β= −0.109, p = 0.008, 95% CI = −0.189 to −0.029), NCOA6 (β= −0.077, p = 0.001, 95% CI = −0.122 to −0.032), and FOXJ3 (β= −0.190, p = 0.002, 95% CI = −0.308 to −0.072) were protective factors (Table 1). In the funnel plot (Figure S1), the SNPs associated with the candidate genes were found to be largely symmetrically distributed, suggesting that the MR analysis adheres to the principles of the second law of mendelian randomization. The outcomes of the scatter plot (Figure S2) and forest plot (Figure S3) further confirmed that BRD1, CASP2, HNRNPF, and GCC1 act as risk genes, whereas BTN2A1, MSL2, NCOA6, and FOXJ3 serve as protective genes. The p-value for the IVW heterogeneity test of Cochran’s Q for the 8 candidate genes was > 0.05 (Table 2), and except for CASP2 and HNRNPF, the p-value for the horizontal pleiotropy test of MR Egger for the remaining 6 candidate genes was also > 0.05 (Table 3). The LOO analysis was conducted for these remaining 6 candidate genes and the outcomes confirmed the reliability and stability of the MR analysis. (Figure S4). To verify that the results of the forward analysis were not confounded by reverse causal effects, the Steiger test was performed. The Steiger test for the 6 candidate genes all returned true (Table 4), and thus, these candidate genes were defined as candidate signature genes.
 |
Table 1 Mendelian Randomization Analytic Result
|
 |
Table 2 Results of the Heterogeneity Test
|
 |
Table 3 Results of the Horizontal Pleiotropy Test
|
 |
Table 4 Results of Steiger Test
|
After obtaining the candidate signature genes, LASSO regression and Boruta were utilized to identify candidate biomarkers. In the LASSO analysis, we retained five candidate signature genes (Figure 3A and B), and in the Boruta analysis, we retained six candidate signature genes (Figure 3C). By taking the intersection of these two results, we obtained five candidate biomarkers, which were BRD1, BTN2A1, MSL2, GCC1, NCOA6 and FOXJ3 (Figure 3D). Then, we proceeded to verify the expression levels of five candidate biomarkers within GSE65682 and GSE95233. Only BRD1, MSL2, NCOA6, and FOXJ3 exhibited significant decrease in expression levels in septic samples, with consistent trends across both cohorts (Figure 3E). The AUC for BRD1 and FOXJ3 were 0.940 and 0.953 in the training cohort and in the validation cohort, they were 0.946 and 0.816, consequently, BRD1 and FOXJ3 were as biomarkers for subsequent analysis (Figure 3F).
 |
Figure 3 Screening of key genes for diagnostic analysis. (A) Spectrogram of coefficients; (B)The ten-fold cross-validation; (C) Boruta algorithm to screen for characterized genes; (D) Venn plot for identification of key genes; (E) Expression difference of key genes in validation set and training set; (F) Receiver operating characteristic curve of key genes.
|
We evaluated the prognostic value of the biomarkers in sepsis by developing a nomogram based on BRD1 and FOXJ3 (Figure 4A). With a p-value of 0.206 and a mean absolute error (MAE) below 0.05 (Figure 4B), it was demonstrated that the nomogram possesses high-precision in its predictive ability. The net benefit is positive in the decision curve further confirms the nomogram’s predictive accuracy (Figure 4C). Ultimately, the AUC of 0.96 substantiates the value of the nomogram in the prediction of sepsis (Figure 4D). These results indicate that the nomogram constructed based on biomarkers has a high predictive accuracy.
 |
Figure 4 Construction of the nomogram diagram model. (A) Nomogram; (B) Calibration curves for nomogram; (C) Decision curve analysis curve; (D) Receiver operating characteristic curves for nomogram.
|
The composition of immune cells in sepsis and control groups was shown in Figure 5A. We then examined the expression patterns of 22 types of immune cells and found significant differences in naive B cells, resting dendritic cells, CD8 T cells, and 15 other immune cell types (Figure 5B). Correlation analysis among these differentially expressed immune cells revealed significant correlations of various extents within the 18 identified cell types, such as there was a significant negative correlation between B cell memory and naive B cells. Additionally, monocytes show significant negative correlations with gamma delta T cells, macrophages, neutrophils, and CD4 naive T cells (Figure 5C). Lastly, the investigation into the relationships between the biomarkers and the differential immune cells uncovered that BRD1 is significantly negatively correlated with Macrophages M0 (p < 0.001, cor = −0.311) and Macrophages M1 (p < 0.001, cor = −0.304). Similarly, FOXJ3 exhibited a significant negative correlation with Macrophages M0 (p < 0.001, cor = −0.308) (Figure 5D).
 |
Figure 5 Immune infiltration and functional enrichment analysis. (A) Cell abundance map of immune infiltration; (B) Differences in immune cell expression between sepsis and control groups; (C) Correlation heat map of immune cells; (D) Heatmap of correlation between differential immune cells and biomarkers; (E) Gene set enrichment analysis of BRD1; (F) Gene set enrichment analysis of FOXJ3.
|
Afterward, we performed GSEA to investigate the potential pathways related to the biomarkers. BRD1 was found to be enriched in 365 pathways, while FOXJ3 was enriched in 354 pathways. Notably, both BRD1 and FOXJ3 were associated with functions like extracellular matrix organization, keratinization, and mRNA splicing (Figure 5E and F).
Based on the miRNet database, we collectively predicted 562 miRNAs and 982 lncRNAs. Specifically, for FOXJ3, we predicted 513 miRNAs and 956 lncRNAs, and for BRD1, we predicted 189 miRNAs and 789 lncRNAs. There were 26 miRNAs that commonly regulated both BRD1 and FOXJ3. Using these 26 miRNAs as the core, we matched them to 299 lncRNAs and constructed a lncRNA-miRNA-mRNA network with 327 nodes and 1,434 edges (Figure 6A). In the JASPAR database, we identified 14 transcription factors (TFs) that regulate the biomarkers. For FOXJ3, we predicted 11 TFs, and for BRD1, predicted 3 TFs, then constructed a TF-mRNA regulatory network that includes 16 nodes and 14 edges (Figure 6B).
 |
Figure 6 Molecular regulatory networks. (A) lncRNA-miRNA-biomarker network. Orange represents biomarkers, green represents key miRNAs, and yellow represents lncRNAs. (B) TF-biomarker network. Orange represents biomarkers, and blue represents transcription factors (TFs). (C) GGI network. Different colored lines represent different interactions, and different colored blocks represent the potential functions involved.
|
To further identify genes related to the biomarkers, GeneMANIA was used to discover 20 genes associated with the biomarkers. Within the interaction network, physical interactions accounted for the highest proportion (77.64%), followed by co-expression interactions (8.01%) and the biomarkers and the genes that interact with them were functionally associated with acetyltransferase complexes, protein acetyltransferase complexes, and the process of protein acetylation (Figure 6C).
Chromosomal localization analysis results show that FOXJ3 is located on chromosome 1, and BRD1 is located on chromosome 22 (Figure 7A). To identify potential drugs for targeted therapy of the biomarkers, we conducted a search in the Dsigdb database and discovered 18 significant drugs, among which digoxin, doxorubicin and daunorubicin simultaneously target both BRD1 and FOXJ3. Subsequently, an mRNA-drug network consisting of 20 nodes and 24 edges was constructed (Figure 7B).
 |
Figure 7 Chromosome localization and drug prediction results. (A) The localization of biomarkers on chromosomes. The outer circle numbers represent different chromosomes, and the inner circle shows the position of the biomarkers on the chromosomes. (B) Drug-biomarker network diagram. Orange represents biomarkers, and blue represents drugs.
|
After rigorous quality control, 16,921 high-quality cells and 20,696 genes were retained for further analysis (Figure S5a and b). The top 2,000 highly variable genes were selected, and 10 of the most variable genes, including CXCL8, CCL4, HBD, and LCN2, were highlighted (Figure S5c). Principal component analysis (PCA) was then performed on these genes for dimensionality reduction, and the top 30 principal components, based on statistical significance (P < 0.05), were selected (Figure S5d and e). A total of 9 cell populations were ultimately identified (Figure 8A). Cell annotation was performed through marker gene analysis, identifying 9 distinct cell types: T cell, T cell precursors, Spermatogenic cells, B cells, Epithelial cells, Monocytes, Neutrophils, Macrophages, and Erythrocytes (Table 5, Figure 8B). The marker genes CD79A, MS4A1, and CD19 exhibited high expression in B cells (Figure 8C). Furthermore, the key gene FOXJ3 showed significant and high expression in Spermatogenic cells, thus Spermatogenic cells were identified as the key cell type (Figure 8D). Cell communication analysis revealed that the spermatogenic cells had strong interactions with neutrophils, monocytes, and B cells (Figure 8E–H). The key gene FOXJ3 was expressed in spermatogenic cells, and the proposed time-series analysis indicated that the expression of FOXJ3 was low throughout the different differentiation stages of spermatogenic cells. (Figure S6a and b).
 |
Table 5 Cell Annotation Types and Their Corresponding Marker Genes
|
 |
Figure 8 Single-cell analysis results. (A) Results of UMAP cell clustering analysis. (B) Manually annotated cell types. (C) Bubble plot of marker gene expression in each cell cluster. (D) Expression of FOXJ3 in different cell clusters. (E) Number of interactions between different cell clusters. (F) Interaction strength between different cell clusters. (G) Heatmap of the number of interactions between different cell clusters. (H) Heatmap of interaction strength between different cell clusters.
|
As shown in Figure 9, BRD1 was significantly decreased in the blood samples of sepsis patients, which was consistent with the bioinformatics analysis. However, although the expression trend of FOXJ3 was consistent with the bioinformatics analysis, there was no significant difference in the expression levels between sepsis patients and control blood samples.
 |
Figure 9 Experimental verification of biomarkers in sepsis. * P < 0.05 and Not Significant (ns): P > 0.05.
|
Sepsis is a hyper-heterogeneous syndrome, accompanied by systemic inflammatory responses throughout the entire course of the disease. Moreover, the inflammatory and immune responses change dynamically at different pathogenic stages.42 Parthanatos, a form of cell death distinct from apoptosis and necrosis, has been implicated in various diseases, including cancer and neurodegenerative disorders.11 However, few studies exist between sepsis and Parthanatos. In this study, we identified six candidate signature genes-BRD1, BTN2A1, MSL2, GCC1, NCOA6, and FOXJ3-via MR analysis. Afterwards, through machine learning, expression validation and ROC analysis, BRD1 and FOXJ3 were retained as biomarkers for further research. We also explored their roles in immune infiltration, molecular regulatory networks, GSEA, chromosomal distribution, and drug prediction.
Bromodomain protein 1 (BRD1) is a gene associated with several psychiatric disorders.43 It encodes a scaffold protein that interacts with epigenetic modifiers involved in histone acetylation and histone H3 N-tail cleavage, regulating the brain development and in a number of biological processes including cell proliferation, differentiation and development are also crucial.44 BRD1’s association with mental disorders, including depression, epilepsy, and schizophrenia, has been supported by various genetic studies.45 Kerstin Klein et al discovered that the inhibition of BRD1 can reduce the expression of lipopolysaccharide-induced tumor necrosis factor-α (TNF-α), suppress the proliferation of synovial fibroblasts, and exert beneficial effects on rheumatoid arthritis (RA). Thus, BRD1 is regarded as a potential therapeutic target for RA.46 Recent evidence indicates that silencing the BRD1 gene in rheumatoid arthritis (RA) patient-derived macrophages (MDMs) exerts a subtle anti-inflammatory effect. Specifically, BRD1 knockdown reduced TNF-α mRNA expression upon LPS stimulation but did not alter TNF-α-induced TNF-α mRNA expression. This suggests that BRD1-mediated regulation of TNF-α may be stimulus-dependent.46 These findings suggest that BRD1 may modulate the initiation or suppression of pyroptosis in immune cells by regulating key inflammatory mediators such as TNF-α, with effects contingent upon specific stimuli (eg, LPS vs TNF-α). We further propose that in the context of sepsis, BRD1 likely operates through distinct inflammatory signaling cascades. Notably, during bacterial infections (exemplified by LPS-induced inflammation), BRD1 may attenuate maladaptive inflammatory responses by downregulating TNF-α expression.Both sepsis and rheumatoid arthritis (RA) are inflammatory diseases. It is also possible that BRD1 could serve as a therapeutic target for sepsis by reducing the expression of inflammatory factors. Functionally, BRD1 plays a role in mitochondrial bioenergetics, impacting mitochondrial function.47 It is worth noting that, according to reports, multiple factors during sepsis can cause mitochondrial damage. The damaged mitochondria actively participate in the formation of the inflammatory environment through key signaling pathways (such as Toll-like receptors), which exacerbates the occurrence of sepsis.48 Therefore, it is speculated that BRD1 can affect mitochondria and thus be involved in the occurrence and development of sepsis. Mitochondrial-targeted intervention may be a prospective target for the treatment of sepsis. To date, no association between BRD1 and sepsis has been reported in the literature, and it has been found for the first time that we will continue to investigate its role in septic diseases and aim to elucidate its underlying mechanisms in future studies.
FOXJ3, a transcription factor belonging to the Forkhead box (FOX) family, is involved in processes such as cell proliferation, migration, and spermatogenesis. The Role of FOX Family Member FOXJ1 in Regulating T Cell Activation and AutoreactivityFOXJ1, a member of the FOX family, plays a role in regulating T cell activation and autoreactivity. The deficiency of this gene in T cells can lead to multi-organ systemic inflammation and an increase in NF-κB activity. Therefore, FOXJ1 may regulate the inflammatory response and prevent autoimmunity by antagonizing pro-inflammatory transcriptional activities.49 In addition, a cross-sectional study has revealed that seven single nucleotide polymorphisms (SNPs), including rs2455084, rs1393009, and rs7539485, in the FOXJ3 gene are significantly associated with the susceptibility to rheumatoid arthritis (RA). The polymorphisms of FOXJ3 are related to the diagnostic indicators reflecting the disease activity of RA.50 Although there is currently no direct evidence linking FOXJ3 to sepsis, previous studies have demonstrated that both FOXJ1 and FOXJ3 are closely associated with inflammatory responses. This suggests that FOXJ3 may participate in the pathogenesis and progression of sepsis by modulating immune responses, particularly through its effects on T-cell function. Notably, recent studies have revealed that FOXJ3 knockout ameliorates arthritis symptoms, further supporting its role in inflammatory regulation.51 Based on these findings, we hypothesize that FOXJ3 may attenuate excessive immune activation in sepsis by antagonizing pro-inflammatory transcription factors, thereby modulating systemic inflammatory responses. Although a direct association between FOXJ3 and sepsis remains to be established, these insights identify FOXJ3 as a potential therapeutic target and provide a novel perspective for investigating the immune mechanisms underlying sepsis pathogenesis.
The effectiveness, sensitivity, and specificity of biomarkers are typically evaluated using the area under the curve (AUC). Both BRD1 and FOXJ3 demonstrated excellent diagnostic performance with AUC values exceeding 0.8 in both training and validation cohorts.We identified BRD1 (AUC 0.946) as a significant diagnostic biomarker for sepsis, outperforming procalcitonin (PCT, AUC 0.79).52 Furthermore, FOXJ3 achieved an AUC of 0.816, surpassing the current gold standard for immunoparalysis detection – monocyte HLA-DR (AUC 0.73)53 expression.Given the current lack of direct multicenter comparative data for these novel targets, future prospective validation across different cohorts is warranted. Combined analysis with established biomarkers (eg, PCT, IL-6) could further confirm their diagnostic value.
Both BRD1 and FOXJ3 were found to be associated with pathways involved in extracellular matrix organization and mRNA splicing. The extracellular matrix is a complex network of macromolecules that play a critical role in cellular adhesion, migration, proliferation, and differentiation.54 During sepsis, inflammatory factors stimulate the degradation of the extracellular matrix, particularly collagen and elastin, leading to remodelling and loss of function mRNA splicing, the process by which introns are removed and exons are joined to form mature mRNA, is fundamental to gene expression regulation.55 Dysregulation of this process is linked to various genetic disorders, including cancers, neurodegenerative diseases, and muscular disorders.56 In sepsis, abnormal mRNA splicing not only affects the balance of the immune response, but may also exacerbate organ damage and influence the course of the disease An in-depth study of these signalling pathways could help with the development of new therapeutic strategies for the treatment of sepsis.
BRD1 exhibited a significant negative correlation with Macrophages M0 and Macrophages M1, while FOXJ3 showed a significant negative correlation with Macrophages M0. M0 macrophages represent an unpolarized naive state derived from bone marrow monocytes, possessing low-level antigen-presenting capacity and basal secretory activity of inflammatory mediators. Under microenvironmental stimulation, they can polarize into either M1 or M2 phenotypes, where M1 macrophages primarily mediate pro-inflammatory responses while M2 macrophages exert anti-inflammatory effects.51 During the early phase of sepsis, M0 macrophages rapidly differentiate into the M1 phenotype, triggering systemic inflammatory response syndrome (SIRS). In later stages, they may differentiate into M2 macrophages or remain in the M0 state, leading to immunoparalysis and increased infection risk.BRD1 may epigenetically suppress the TLR4/NF-κB pathway, thereby reducing the propensity of M0-to-M1 conversion, whereas FOXJ3 may negatively regulate IRF5 – a critical determinant of M1 polarization – to maintain macrophages in the M0 state. Together, these factors form a “molecular brake” for M0 homeostasis, whose dysregulation represents a key contributor to immune imbalance in sepsis.Targeted regulation of the macrophage polarization process may normalize the host immune response, thereby providing a new therapeutic approach for the treatment of sepsis.57 Macrophage polarization likely serves as a central target for immune modulation in sepsis. As the differentiation origin, M0 macrophages hold upstream regulatory value for intervention strategies. Compared to other inflammatory cells such as neutrophils, macrophages demonstrate superior plasticity and greater accessibility for both research and therapeutic purposes.Furthermore, Jamie R Privratsky et al have discovered that the macrophage-endothelial immunomodulatory axis is also a crucial regulatory axis for ameliorating acute lung injury in sepsis.58 Therefore, it is speculated that biomarkers (such as BRD1 and FOXJ3) may participate in the occurrence and development of sepsis by influencing the infiltration status of macrophages.Our findings demonstrate significant differences in regulatory T cell (Treg) abundance between the sepsis and control groups. Tregs play a crucial role in mitigating excessive inflammation during early-stage sepsis, while their upregulated frequency in late stages may induce immunosuppression.59 BRD1 potentially upregulates gene expression associated with immunosuppressive cells (including Tregs), leading to abnormal increases in both Treg quantity and function. These hyperplastic Tregs may suppress the activation and proliferation of effector T cells and natural killer cells, thereby impairing pathogen clearance capacity.The suppressive function of Tregs largely depends on high-level and stable expression of the transcription factor FOXP3.60 However, excessive FOXP3 activation may also cause abnormal Treg expansion, consequently weakening the host’s immune response against infections. Therefore, further investigation of BRD1 and FOXJ3 in immune regulation is warranted, which would help elucidate sepsis-related immune mechanisms and identify novel potential therapeutic targets for sepsis treatment.
Common Targets for BRD1 and FOXJ3: the drug prediction analysis identified digoxin, doxorubicin, and daunorubicin as common drugs targeting both BRD1 and FOXJ3. Digoxin, a cardiac glycoside derived from the foxglove plant, is primarily used to treat heart conditions and has potential preventive effects in osteoarthritis and intervertebral disc degeneration.61,62 Doxorubicin, an anthracycline antibiotic derived from streptomyces peucetius, has been widely used as a chemotherapeutic agent since the 1960s.63 It is reported that sialic acid-conjugate modified doxorubicin can treat neutrophil-related inflammation.64,65 However, studies have demonstrated that both digoxin and doxorubicin exhibit certain toxicity, particularly potentially inducing cardiotoxicity at excessive doses.66 Daunorubicin, another anthracycline antibiotic, is used to treat a variety of cancers, particularly leukemia. Similar to doxorubicin, it intercalates into DNA, inhibiting DNA and RNA synthesis to exert its antitumor effects.67 Given the therapeutic potential and safety concerns of these drugs in sepsis treatment, future studies should establish animal models to thoroughly investigate the mechanisms of action and therapeutic effects of digoxin and doxorubicin in sepsis. Through animal experiments, we can evaluate the pharmacological responses of these drugs during different sepsis stages, particularly their impacts on immune and inflammatory responses.The pathogenesis of sepsis involves complex and diverse mechanisms, and the prediction and development of related drugs still require further exploration in future research. As potential biomarkers, BRD1 and FOXJ3 could not only facilitate early sepsis diagnosis but also provide evidence for personalized treatment strategies.Precision diagnosis and drug prediction based on these biomarkers may offer novel therapeutic targets and approaches for sepsis patients in the future. This could lead to improved treatment outcomes, reduced medication risks, and advancement of personalized medicine.The pathogenesis of sepsis is complex and varied, and the prediction and development of its drugs still need to be explored in future research.
This study also has limitations in that the study was analyzed by MR, which has limitations, such as the validity and applicability of genetic tools and the possibility of genetic heterogeneity, which should be carefully considered when performing MR studies. Furthermore, the RT-qPCR results demonstrated that BRD1 was consistent with bioinformatics findings, confirming its potential as a biomarker for sepsis. Although FOXJ3 showed a similar expression trend, it did not exhibit significant differences between disease and healthy blood samples. This discrepancy may be due to the small sample size, batch effects, or differences in sample sources, necessitating further studies to explore its role in disease.
In summary, in this study, based on the public data download transcriptomics data and parthanatos-related genes, we highlight BRD1 and FOXJ3 as promising biomarkers for sepsis through a number of columns of bioinformatics methods. In addition, we performed pathway enrichment analysis, immune cell infiltration status and drug prediction of these genes, which provided new insights into the management of sepsis. Future studies will expand sample sizes and employ BRD1 knockout mouse models or RNA interference technology to investigate the role of BRD1 in programmed cell death within immune cells. Optimized disease models will be established to evaluate clinically relevant outcomes and compare them with in vitro experimental results, thereby validating the robustness of conclusions and laying the foundation for translational applications.The investigation will focus on elucidating the functions of these genes and related pathways to provide a theoretical basis for developing targeted therapies. Additionally, animal models will be utilized to assess the safety, toxicity, and optimal therapeutic windows of digoxin and doxorubicin, generating evidence for clinical applications. These studies will further verify the drugs’ specific effects on BRD1/FOXJ3, offering theoretical support for advancing targeted therapeutic strategies.Follow-up studies will further focus on the role of these genes and related pathways to provide a theoretical basis for the development of targeted therapies.
PRGs, Parthanatos-related genes; DEGs, Differentially expressed genes; MR, Mendelian randomization; ROC, Receiver operating characteristic; RT-qPCR, Real-time quantitative polymerase chain reaction; NAD, Nicotinamide adenine dinucleotide; ATP, Adenosine triphosphate; AIF, Apoptosis inducing factor; SNPs, Single nucleotide polymorphisms; GEO, Gene expression omnibus; PRGs, Parthanatos-related genes; WGCNA, Weighted gene co-expression network analysis; GO, Gene ontology; KEGG, Kyoto encyclopedia of genes and genomes; PPI, Protein-protein interaction; IV, Instrumental variables; IVW, Inverse variance weighted; LASSO, Least absolute shrinkage and selectionoperator; AUC, Area under the curve; TFs, Transcription factors; MF, Molecular functions; BP, Biological processes; CC, Cellular components; BRD1, Bromodomain protein 1; TNF-α, Tumor necrosis factor-α; RA, Rheumatoid arthritis.
The original data presented in the study are openly available in [Gene Expression Omnibus] at [GSE65682, GSE95233, http://www.ncbi.nlm.nih.gov/geo/], [Integrative Epidemiology Unit (IEU) Open Genome-Wide Association Study] at [ieu-b-4980, https://gwas.mrcieu.ac.uk/].
I certify that the research study titled Transcriptome Combined with Mendelian Randomization to Identify and Validate Biomarkers Associated with Parthanatos in Sepsis has been approved by the Shanxi Bethune Hospital Ethics Committee.The approval number and date of approval are as follows: YXLL-2023-269 and 2023.12.8.
We would like to express our sincere gratitude to all individuals and organizations who supported and assisted us throughout this research. Special thanks to the following authors: Suojuan Zhang and Jifang Liang. In conclusion, we extend our thanks to everyone who has supported and assisted us along the way. Without your support, this research would not have been possible.
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
This work was supported by Clinical Key Specialty Program of Shanxi Bethune Hospital and the Shanxi Provincial Department of Science and Technology,Shanxi Provincial Basic Research Program, Free Exploration Youth Scientific Research Project [grant number 202403021212186].
The authors declare no conflicts of interest in this work. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.
1. Piva S, Bertoni M, Gitti N, Rasulo FA, Latronico N. Neurological complications of sepsis. Curr Opin Crit Care. 2023;29(2):75–84. doi:10.1097/MCC.0000000000001022
2. Fleischmann-Struzek C, Mellhammar L, Rose N, et al. Incidence and mortality of hospital- and ICU-treated sepsis: results from an updated and expanded systematic review and meta-analysis. Intensive Care Med. 2020;46(8):1552–1562. doi:10.1007/s00134-020-06151-x
3. Rudd KE, Johnson SC, Agesa KM, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200–211. doi:10.1016/S0140-6736(19)32989-7
4. Qiu X, Lei YP, Zhou RX. SIRS, SOFA, qSOFA, and NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther. 2023;21(8):891–900. doi:10.1080/14787210.2023.2237192
5. Llitjos JF, Carrol ED, Osuchowski MF, et al. Enhancing sepsis biomarker development: key considerations from public and private perspectives. Crit Care. 2024;28(1):238. doi:10.1186/s13054-024-05032-9
6. Zampieri FG, Bagshaw SM, Semler MW. Fluid therapy for critically ill adults with sepsis: a review. JAMA. 2023;329(22):1967–1980. doi:10.1001/jama.2023.7560
7. Mirijello A, Tosoni A, Group OBOTIMSS. New strategies for treatment of sepsis. Medicina. 2020;56(10):527. doi:10.3390/medicina56100527
8. Huang P, Chen G, Jin W, Mao K, Wan H, He Y. Molecular mechanisms of parthanatos and its role in diverse diseases. Int J Mol Sci. 2022;23(13).
9. Koehler RC, Dawson VL, Dawson TM. Targeting parthanatos in ischemic stroke. Front Neurol. 2021;12:662034. doi:10.3389/fneur.2021.662034
10. Morana O, Wood W, Gregory CD. The apoptosis paradox in cancer. Int J Mol Sci. 2022;23(3):1328. doi:10.3390/ijms23031328
11. Elmore S. Apoptosis: a review of programmed cell death. Toxicol Pathol. 2007;35(4):495–516. doi:10.1080/01926230701320337
12. Lorente L, Martín MM, Ortiz-López R, et al. Parthanatos type programmed cell death and septic patient mortality. Medicina intensiva. 2023;47(12):691–696. doi:10.1016/j.medin.2023.04.016
13. Duan L, Xiao R, Liu S, Shi Y, Feng Y. Causality between cognitive performance and cardiovascular disease: a bidirectional Mendelian randomization study. Gene. 2024;891:147822. doi:10.1016/j.gene.2023.147822
14. Sekula P, M FDG, Pattaro C, Köttgen A. Mendelian randomization as an approach to assess causality using observational data. J Am Soc Nephrol. 2016;27(11):3253–3265. doi:10.1681/ASN.2016010098
15. Burgess S, Thompson SG, Collaboration CCG. Avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol. 2011;40(3):755–764. doi:10.1093/ije/dyr036
16. Fan S, Li H, Liu K. Molecular prognostic of nine parthanatos death-related genes in glioma, particularly in COL8A1 identification. J Neurochem. 2024;168(3):205–223. doi:10.1111/jnc.16049
17. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. doi:10.1093/nar/gkv007
18. Gustavsson EK, Zhang D, Reynolds RH, Garcia-Ruiz S, Ryten M. ggtranscript: an R package for the visualization and interpretation of transcript isoforms using ggplot2. Bioinformatics. 2022;38(15):3844–3846. doi:10.1093/bioinformatics/btac409
19. Gu Z, Hübschmann D. Make Interactive Complex Heatmaps in R. Bioinformatics. 2022;38(5):1460–1462. doi:10.1093/bioinformatics/btab806
20. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinf. 2008;9:559. doi:10.1186/1471-2105-9-559
21. Zheng Y, Gao W, Zhang Q, et al. Ferroptosis and autophagy-related genes in the pathogenesis of ischemic cardiomyopathy. Front Cardiovasc Med. 2022;9:906753. doi:10.3389/fcvm.2022.906753
22. Yu G, Wang L-G, Han Y, He Q-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics. 2012;16(5):284–287. doi:10.1089/omi.2011.0118
23. Liu P, Xu H, Shi Y, Deng L, Chen X. Potential molecular mechanisms of plantain in the treatment of gout and hyperuricemia based on network pharmacology. Evidence-Based Complementary Alternative Med. 2020;2020:3023127. doi:10.1155/2020/3023127
24. Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife. 2018;7.
25. Xian W, Wu D, Liu B, et al. Graves disease and inflammatory bowel disease: a bidirectional Mendelian randomization. J Clin Endocrinol Metab. 2023;108(5):1075–1083. doi:10.1210/clinem/dgac683
26. Liu Z, Wang H, Yang Z, Lu Y, Zou C. Causal associations between type 1 diabetes mellitus and cardiovascular diseases: a Mendelian randomization study. Cardiovasc Diabetol. 2023;22(1):236. doi:10.1186/s12933-023-01974-6
27. Mounier N, Kutalik Z. Bias correction for inverse variance weighting Mendelian randomization. Genetic Epidemiol. 2023;47(4):314–331. doi:10.1002/gepi.22522
28. Xu J, Zhang S, Tian Y, et al. Genetic Causal Association between iron status and osteoarthritis: a two-sample Mendelian Randomization. Nutrients. 2022;14(18):3683. doi:10.3390/nu14183683
29. Dai Z, Xu W, Ding R, et al. Two-sample Mendelian randomization analysis evaluates causal associations between inflammatory bowel disease and osteoporosis. Front Public Health. 2023;11:1151837. doi:10.3389/fpubh.2023.1151837
30. Verbanck M, Chen C-Y, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018;50(5):693–698. doi:10.1038/s41588-018-0099-7
31. Li Y, Lu F, Yin Y. Applying logistic LASSO regression for the diagnosis of atypical Crohn’s disease. Sci Rep. 2022;12(1):11340. doi:10.1038/s41598-022-15609-5
32. Yue S, Li S, Huang X, et al. Machine learning for the prediction of acute kidney injury in patients with sepsis. J Transl Med. 2022;20(1):215. doi:10.1186/s12967-022-03364-0
33. Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinf. 2011;12:77. doi:10.1186/1471-2105-12-77
34. Sachs MC. plotROC: a tool for plotting ROC curves. J Statistical Software. 2017;79. doi:10.18637/jss.v079.c02
35. Vickers AJ, Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res. 2019;3:18. doi:10.1186/s41512-019-0064-7
36. Hu Y, Yan C, Hsu C-H, et al. OmicCircos: a simple-to-use R package for the circular visualization of multidimensional omics data. Cancer Inf. 2014;13:13–20. doi:10.4137/CIN.S13495
37. Kuleshov MV, Jones MR, Rouillard AD, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res. 2016;44(W1):W90–7. doi:10.1093/nar/gkw377
38. Hao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573–87.e29. doi:10.1016/j.cell.2021.04.048
39. Li Y, Yu J, Li R, Zhou H, Chang X. New insights into the role of mitochondrial metabolic dysregulation and immune infiltration in septic cardiomyopathy by integrated bioinformatics analysis and experimental validation. Cell Mol Biol Lett. 2024;29(1):21. doi:10.1186/s11658-024-00536-2
40. Jin S, Guerrero-Juarez CF, Zhang L, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12(1):1088. doi:10.1038/s41467-021-21246-9
41. Qiu X, Hill A, Packer J, Lin D, Ma YA, Trapnell C. Single-cell mRNA quantification and differential analysis with Census. Nat Methods. 2017;14(3):309–315. doi:10.1038/nmeth.4150
42. Rubin J, Berra L. Electrical impedance tomography in the adult intensive care unit: clinical applications and future directions. Curr Opin Crit Care. 2022;28(3):292–301. doi:10.1097/MCC.0000000000000936
43. Rajkumar AP, Qvist P, Donskov JG, et al. Reduced Brd1 expression leads to reversible depression-like behaviors and gene-expression changes in female mice. Transl Psychiatry. 2020;10(1):239. doi:10.1038/s41398-020-00914-2
44. Paternoster V, Edhager AV, Qvist P, et al. Inactivation of the Schizophrenia-associated BRD1 gene in brain causes failure-to-thrive, seizure susceptibility and abnormal histone H3 acetylation and N-tail clipping. Mol Neurobiol. 2021;58(9):4495–4505. doi:10.1007/s12035-021-02432-8
45. Paternoster V, Svanborg M, Edhager AV, et al. Brain proteome changes in female Brd1± mice unmask dendritic spine pathology and show enrichment for schizophrenia risk. Neurobiol Dis. 2019;124:479–488. doi:10.1016/j.nbd.2018.12.011
46. Klein K, Kato M, Frank-Bertoncelj M, et al. Evaluating the bromodomain protein BRD1 as a therapeutic target in rheumatoid arthritis. Sci Rep. 2018;8(1):11125. doi:10.1038/s41598-018-29127-w
47. Paternoster V, Cömert C, Kirk LS, et al. The psychiatric risk gene BRD1 modulates mitochondrial bioenergetics by transcriptional regulation. Transl Psychiatry. 2022;12(1):319. doi:10.1038/s41398-022-02053-2
48. Hu D, Sheeja Prabhakaran H, Zhang YY, Luo G, He W, Liou YC. Mitochondrial dysfunction in sepsis: mechanisms and therapeutic perspectives. Crit Care. 2024;28(1):292. doi:10.1186/s13054-024-05069-w
49. Lin L, Spoor MS, Gerth AJ, Brody SL, Peng SL. Modulation of Th1 activation and inflammation by the NF-kappaB repressor Foxj1. Science. 2004;303(5660):1017–20.
50. Ban JY, Park HJ, Kim SK, et al. Association of forkhead box J3 (FOXJ3) polymorphisms with rheumatoid arthritis. Mol Med Rep. 2013;8(4):1235–1241. doi:10.3892/mmr.2013.1623
51. Yuan L, Jiang N, Li Y, Wang X, Wang W. RGS1 enhancer RNA promotes gene transcription by recruiting transcription factor FOXJ3 and facilitates osteoclastogenesis through PLC-IP3R-dependent Ca(2+) response in rheumatoid arthritis. Inflammation. 2025;48(1):447–463. doi:10.1007/s10753-024-02067-6
52. Ito A, Ishida T, Tachibana H, Arita M, Yamazaki A, Washio Y. Utility of procalcitonin for differentiating cryptogenic organising pneumonia from community-acquired pneumonia. Clin Chem Lab Med. 2019;57(10):1632–1637.
53. Mosayebi G, Alizadeh SA, Alasti A, et al. Is CD19 an immunological diagnostic marker for acute appendicitis? Iran J Immunol. 2013;10(4):216–228.
54. Karamanos NK, Theocharis AD, Piperigkou Z, et al. A guide to the composition and functions of the extracellular matrix. FEBS J. 2021;288(24):6850–6912. doi:10.1111/febs.15776
55. Rogalska ME, Vivori C, Valcárcel J. Regulation of pre-mRNA splicing: roles in physiology and disease, and therapeutic prospects. Nat Rev Genet. 2023;24(4):251–269. doi:10.1038/s41576-022-00556-8
56. Nikom D, Zheng S. Alternative splicing in neurodegenerative disease and the promise of RNA therapies. Nat Rev Neurosci. 2023;24(8):457–473. doi:10.1038/s41583-023-00717-6
57. Chen X, Liu Y, Gao Y, Shou S, Chai Y. The roles of macrophage polarization in the host immune response to sepsis. Int Immunopharmacol. 2021;96:107791. doi:10.1016/j.intimp.2021.107791
58. Privratsky JR, Ide S, Chen Y, et al. A macrophage-endothelial immunoregulatory axis ameliorates septic acute kidney injury. Kidney Int. 2023;103(3):514–528. doi:10.1016/j.kint.2022.10.008
59. Huang S, Liu D, Han L, et al. Decoding the potential role of regulatory T cells in sepsis-induced immunosuppression. Eur J Immunol. 2024;54(5):e2350730. doi:10.1002/eji.202350730
60. Vent-Schmidt J, Han JM, MacDonald KG, Levings MK. The role of FOXP3 in regulating immune responses. Int Rev Immunol. 2014;33(2):110–128. doi:10.3109/08830185.2013.811657
61. Wang K-D, Ding X, Jiang N, et al. Digoxin targets low density lipoprotein receptor-related protein 4 and protects against osteoarthritis. Ann Rheumatic Dis. 2022;81(4):544–555. doi:10.1136/annrheumdis-2021-221380
62. Meng Q, Liu K, Liu Z, et al. Digoxin protects against intervertebral disc degeneration via TNF/NF-κB and LRP4 signaling. Front Immunol. 2023;14:1251517. doi:10.3389/fimmu.2023.1251517
63. Carvalho C, Santos RX, Cardoso S, et al. Doxorubicin: the good, the bad and the ugly effect. Curr Med Chem. 2009;16(25):3267–3285. doi:10.2174/092986709788803312
64. Wang S, Lai X, Li C, et al. Sialic acid-conjugate modified doxorubicin nanoplatform for treating neutrophil-related inflammation. J Control Release. 2021;337:612–627. doi:10.1016/j.jconrel.2021.07.044
65. Peter S, Alven S, Maseko RB, Aderibigbe BA. Doxorubicin-based hybrid compounds as potential anticancer agents: a review. Molecules. 2022;27(14):4478. doi:10.3390/molecules27144478
66. Patocka J, Nepovimova E, Wu W, Kuca K. Digoxin: pharmacology and toxicology-A review. Environ Toxicol Pharmacol. 2020;79:103400. doi:10.1016/j.etap.2020.103400
67. Chiou J-T, Huang C-H, Wu T-H, et al. CREB/Sp1-mediated MCL1 expression and NFκB-mediated ABCB1 expression modulate the cytotoxicity of daunorubicin in chronic myeloid leukemia cells. Toxicol Appl Pharmacol. 2022;435:115847. doi:10.1016/j.taap.2021.115847
At least 176 people lost their lives and several remained missing as flash floods wreaked havoc across Khyber Pakhtunkhwa on Friday, according to data from the Provincial Disaster Management Agency (PDMA).
Since late June, monsoon rains have wreaked havoc across the country — especially KP and northern regions — by triggering deadly floods, landslides and displacement, particularly in vulnerable, poorly drained, or densely populated areas.
The province-wide deaths included 150 men, 14 women and 12 children, with Buner witnessing the highest number of deaths, 78, according to the PDMA. The data added that 45 homes, three schools and eight other structures were destroyed amid the deluge, with 26 homes being destroyed in Swat alone.
Earlier, the KP government dispatched a helicopter with supplies to Bajaur district, but said that it lost contact with the aircraft. It later confirmed in a statement that the helicopter had crashed and five passengers were killed.
“As a result of this tragic accident, five passengers, including two pilots, were martyred,” the statement read, quoting KP Chief Minister Ali Amin Gandapur.
“The provincial government has called for a day of mourning tomorrow and flags will be flown at half mast,” the statement added. “Rescue teams have been dispatched to the crash site and the martyrs will be buried with full honours.”
In an earlier statement, the CM was quoted as saying that contact with the helicopter had been lost due to “bad weather”.
Buner Deputy Commissioner Kashif Qayum Khan told Dawn.com that 78 people had lost their lives, while “several” were missing. A PDMA daily situation report seen by Dawn.com confirmed the casualties, with 75 men, two women and a child among the deceased.
He added that an emergency has been declared across the district as relief efforts continue in flood-affected areas.
“Helicopters are being used to carry out rescue operations in remote and inaccessible regions,” he said, adding that in Pir Baba Bazaar and the adjoining neighbourhood, floodwaters have completely submerged the area.
“A mosque in Gokand was destroyed and a large number of livestock perished,” he added. “Several roads remain blocked, and the exact number of missing persons has yet to be confirmed.”
Officials said the true figure would only be known once floodwaters receded.
Other most-impacted districts included Bajaur — located in the same Malakand Division as Buner — where eight children were among 21 killed and eight were injured due to flash floods, the PDMA report said.
Incidents related to lightning strikes took the lives of 15 men in Battagram, while 14 deaths and two injuries were reported in Mansehra due to floods.
In Swat, flash floods and thunder strikes claimed 11 lives, the PDMA report added. A roof collapse left five men dead and three wounded in Lower Dir, while two men were killed and as many were injured in a similar incident in Shangla.
The KP government said a provincial govt MI-17 rescue helicopter had reached Buner to evacuate people to safe areas.
Meanwhile, Muhammad Sohail, a media coordinator for Rescue 1122, told Dawn.com that more than 157 bodies have so far been recovered, while over 100 people, including women and children, have been rescued and moved to safe locations.
“The situation is at its worst and rescue operations are continuing in the affected areas, as authorities work to reach stranded residents and provide relief,” he said.
Speaking to Geo News, KP Governor Faisal Karim Kundi responded in the affirmative when asked whether an emergency should be declared. An official notification for that is yet to be issued.
PTI MNA Gohar Ali Khan, who is from Malakand Division’s Buner district, told Geo News: “We have sent rescue teams but reaching the points is also difficult.”
Buner District Police Officer (DPO) also told Dawn.com in an earlier statement that 54 bodies were brought to a Tehsil Headquarters Hospital.
Prime Minister Shehbaz Sharif directed the relevant authorities to accelerate the rescue operation in Battagram district. In a statement, he expressed grief over the deaths and prayed for those who lost their lives in the flash flood.
KP Chief Minister Ali Amin Gandapur spoke to the Hazara commissioner and Battagram DC on the phone and directed that the district administration officials reach the site to supervise the rescue operations, his government said.
According to the Associated Press of Pakistan, Battagram Assistant Commissioner Muhammad Saleem Khan said the casualties occurred after five houses were destroyed last night due to a lightning strike in Neel Band village, which is located on the border of Battagram and Mansehra districts.
In Bajaur earlier today, there were “reports of several people injured in flash floods”, which were caused by a cloudburst (heavy rainfall) in Salarzai tehsil’s Jabrarai village“, Rescue 1122 spokesperson Bilal Ahmad Faizi told Dawn.com.
“Rescue 1122 personnel, with the cooperation of residents, have so far recovered 16 bodies and rescued three injured from the rubble and rainwater,” Faizi confirmed, stating an earlier toll.
A search and rescue operation was underway under the supervision of Bajaur District Emergency Officer Amjad Khan as seven people remained missing, Faizi said, citing locals. DEO Amjad Khan and the station house in-charge were personally supervising the operation, the Rescue 1122 official added.
The deluge in Battagram affected villages located on the border areas of Neel Band, Sarim and Malkal Gali, according to a statement issued by Battagram Rescue 1122 spokesperson Aziz Khan.
“The ongoing rescue efforts are facing challenges due to intermittent rain and a near-total loss of mobile network coverage, severely impacting communication,” the statement explained.
In Lower Dir, five people died and four were wounded when the roof of a house in the Maidan area’s Suri Pao village collapsed due to heavy rain, Faizi said.
Detailing the hurdles, the rescue official said: “The rescue team reached the scene after walking for three hours despite heavy rain, flooded rails, difficult and bad roads.”
Yesterday, over a dozen people were killed as rains and flooding ripped through the country’s northern parts, including Azad Jammu and Kashmir (AJK) and Gilgit-Baltistan (GB).
In Muzaffarabad, a massive landslide in Sarli Sacha village hit a home, leaving six members of a family buried and feared dead. Torrential rains claimed the lives of two more women in AJK’s Bagh and Sudhnoti districts.
In GB, flash floods killed at least eight people, with two still missing in the Ghizer district, while also devastating villages in the Khalti, Ishkoman and Yasin areas.
Similarly, a spell of heavy downpour lashed various parts of Abbottabad district yesterday, triggering flash floods that severely disrupted traffic flow and caused damage to infrastructure.
At least 325 people, including 142 children, have died and 743 others have been injured since June 26 in flash floods and torrential rains that have battered several parts of Pakistan, according to daily data from the National Disaster Management Authority (NDMA).
More to follow
ZEUS represents a breakthrough in zero-emission power generation, leveraging high-pressure oxyfuel-based thermal technology with integrated CO₂ capture. The FEED phase aims to demonstrate ZEUS’ ability to generate dispatchable, zero-emission power using high-CO₂ natural gas streams, transforming previously stranded gas reserves into sustainable energy sources. Once validated, the technology may be scaled up for both onshore and offshore deployment, including floating power applications.
Zahid Osman, President & Group CEO of MISC, said “The ZEUS project demonstrates how we are putting our #deliveringProgress mission into action, harnessing innovation to deliver more energy with less emissions. By integrating cutting-edge carbon capture technology with adaptable maritime solutions, we are not just advancing sustainable power generation but also unlocking new possibilities for stranded gas resources. The signing marks a progressive step forward for the team and exemplifies how partnerships can turn bold ideas into impactful solutions. MISC is proud to collaborate and leverage our maritime expertise to advance this technology for offshore deployment, supporting both Malaysia’s and the world’s transition to a lower-carbon future.”
Vice President of Group Technology & Commercialisation, Izwan Ismail, “ZEUS represents more than a technology demonstration — it is a bold statement of PETRONAS’ unwavering commitment to a low-emission, sustainable future. Beyond its technical capabilities, ZEUS fosters innovation ecosystems and expedites the delivery of sustainable solutions, awakening possibilities for our upstream business. We are collaborating with the right partners to accelerate the delivery of this solution.”
Jo Kjetil Krabbe, Executive Vice President of Power Solutions at Aker Solutions, emphasized the importance of robust engineering and effective system integration as the project moves into its next phase. “We look forward to turning proven components into an integrated system that works at scale, and with zero emissions. Aker Solutions’ role is to ensure that the design is technically robust, that interfaces are well managed, and that the system can be built and operated reliably. Together with PETRONAS Research and MISC Berhad, we’re taking an important step toward making ZEUS ready for deployment in demanding operational environments.”
Mark McGough, CEO of Clean Energy Systems said, “The ZEUS project is another validation of the unique enabling capabilities of CES’ rocket-engine based technology, which can be fueled by a wide range of gases. Our proprietary platelet burner technology has been developed and extensively tested for the past two decades, and now CES is excited to bring these new oxy-combustion products to benefit the oil & gas market.”
The partners share a clear ambition: to deliver sustainable, and affordable zero-emission power solutions that support the global energy transition.
The ZEUS Demo FEED is designed to replicate the thermodynamic conditions of a full-scale commercial plant, minimising redesign efforts for future scale-up. Testing will be done in Malaysia to validate the system’s feasibility, with the FEED phase setting the stage for a final investment decision on the Engineering, Procurement, and Construction (EPC) phase in 2026.