Blog

Using artificial intelligence, MIT researchers have come up with a new way to design nanoparticles that can more efficiently deliver RNA vaccines and other types of RNA therapies.

After training a machine-learning model to analyze thousands of existing delivery particles, the researchers used it to predict new materials that would work even better. The model also enabled the researchers to identify particles that would work well in different types of cells, and to discover ways to incorporate new types of materials into the particles.

What we did was apply machine-learning tools to help accelerate the identification of optimal ingredient mixtures in lipid nanoparticles to help target a different cell type or help incorporate different materials, much faster than previously was possible.”

Giovanni Traverso, associate professor of mechanical engineering at MIT, gastroenterologist at Brigham and Women’s Hospital, and senior author of the study

This approach could dramatically speed the process of developing new RNA vaccines, as well as therapies that could be used to treat obesity, diabetes, and other metabolic disorders, the researchers say.

Alvin Chan, a former MIT postdoc who is now an assistant professor at Nanyang Technological University, and Ameya Kirtane, a former MIT postdoc who is now an assistant professor at the University of Minnesota, are the lead authors of the new study, which appears today in Nature Nanotechnology.

Particle predictions

RNA vaccines, such as the vaccines for SARS-CoV-2, are usually packaged in lipid nanoparticles (LNPs) for delivery. These particles protect mRNA from being broken down in the body and help it to enter cells once injected.

Creating particles that handle these jobs more efficiently could help researchers to develop even more effective vaccines. Better delivery vehicles could also make it easier to develop mRNA therapies that encode genes for proteins that could help to treat a variety of diseases.

In 2024, Traverso’s lab launched a multiyear research program, funded by the U.S. Advanced Research Projects Agency for Health (ARPA-H), to develop new ingestible devices that could achieve oral delivery of RNA treatments and vaccines.

“Part of what we’re trying to do is develop ways of producing more protein, for example, for therapeutic applications. Maximizing the efficiency is important to be able to boost how much we can have the cells produce,” Traverso says.

A typical LNP consists of four components – a cholesterol, a helper lipid, an ionizable lipid, and a lipid that is attached to polyethylene glycol (PEG). Different variants of each of these components can be swapped in to create a huge number of possible combinations. Changing up these formulations and testing each one individually is very time-consuming, so Traverso, Chan, and their colleagues decided to turn to artificial intelligence to help speed up the process.

“Most AI models in drug discovery focus on optimizing a single compound at a time, but that approach doesn’t work for lipid nanoparticles, which are made of multiple interacting components,” Chan says. “To tackle this, we developed a new model called COMET, inspired by the same transformer architecture that powers large language models like ChatGPT. Just as those models understand how words combine to form meaning, COMET learns how different chemical components come together in a nanoparticle to influence its properties – like how well it can deliver RNA into cells.”

To generate training data for their machine-learning model, the researchers created a library of about 3,000 different LNP formulations. The team tested each of these 3,000 particles in the lab to see how efficiently they could deliver their payload to cells, then fed all of this data into a machine-learning model.

After the model was trained, the researchers asked it to predict new formulations that would work better than existing LNPs. They tested those predictions by using the new formulations to deliver mRNA encoding a fluorescent protein to mouse skin cells grown in a lab dish. They found that the LNPs predicted by the model did indeed work better than the particles in the training data, and in some cases better than LNP formulations that are used commercially.

Accelerated development

Once the researchers showed that the model could accurately predict particles that would efficiently deliver mRNA, they began asking additional questions. First, they wondered if they could train the model on nanoparticles that incorporate a fifth component: a type of polymer known as branched poly beta amino esters (PBAEs).

Research by Traverso and his colleagues has shown that these polymers can effectively deliver nucleic acids on their own, so they wanted to explore whether adding them to LNPs could improve LNP performance. The MIT team created a set of about 300 LNPs that also include these polymers, which they used to train the model. The resulting model could then predict additional formulations with PBAEs that would work better.

Next, the researchers set out to train the model to make predictions about LNPs that would work best in different types of cells, including a type of cell called Caco-2, which is derived from colorectal cancer cells. Again, the model was able to predict LNPs that would efficiently deliver mRNA to these cells.

Lastly, the researchers used the model to predict which LNPs could best withstand lyophilization – a freeze-drying process often used to extend the shelf-life of medicines.

“This is a tool that allows us to adapt it to a whole different set of questions and help accelerate development. We did a large training set that went into the model, but then you can do much more focused experiments and get outputs that are helpful on very different kinds of questions,” Traverso says.

He and his colleagues are now working on incorporating some of these particles into potential treatments for diabetes and obesity, which are two of the primary targets of the ARPA-H funded project. Therapeutics that could be delivered using this approach include GLP-1 mimics with similar effects to Ozempic.

This research was funded by the GO Nano Marble Center at the Koch Institute, the Karl van Tassel Career Development Professorship, the MIT Department of Mechanical Engineering, Brigham and Women’s Hospital, and ARPA-H.

To date, systematic reviews have been performed to investigate the relationship between periodontal diseases and various health conditions, including cardiovascular diseases, diabetes, complications during pregnancy, and autoimmune disorders [27, 28]. The most recent systematic review assessing the relationship between periodontitis and IBD utilized the same criteria as the present study but focused exclusively on articles published until October 2021 [29]. In the previously referenced study, the researchers discovered that patients with IBD exhibited significantly higher likelihood of developing periodontitis (OR = 2.65). Our recent analysis, which involved studies up to 2024, revealed an OR of 2.28, further supporting this positive correlation (with overlapping CIs).

Consequently, we have undertaken a comprehensive review of pertinent studies exploring the association between periodontitis and IBD, including CD and UC. Both periodontitis and IBD share an inflammatory etiology and are linked to microbial dysbiosis, as well as in certain instances modifications in the host immune response [30]. In total, a comprehensive review of the literature was conducted up to December 2024, including 10 articles for statistical analysis.

Our investigation revealed that patients with IBD are at an increased risk of developing periodontitis compared to individuals without IBD, with an OR of 2.28 (95% CI: 1.73-3). Furthermore, the OR for periodontitis was elevated among patients diagnosed with UC in comparison to those with CD, demonstrating an OR of 3.14 (95% CI: 2.11–4.66) for UC versus an OR of 1.99 (95% CI: 1.4–2.83) for CD. These findings demonstrate the relationship between any form of IBD and the prevalence of periodontitis. In this study, the OR for UC patients was found to be greater than that for CD patients. However, due to the overlapping CIs and a P-value of 0.09 from the formal subgroup difference test, no significant difference was observed between UC and CD. Although this difference was not statistically meaningful, future research into subtype-specific mechanisms such as differential Th1/Th17 polarization or oral-gut microbial translocation is needed.

Owing to stringent inclusion criteria and serious research limitations, not all studies identified for the systematic review were incorporated into the meta-analysis, as evidenced by the findings of the Zervou study [31]. The research involved a clinical examination of 15 patients with UC and 15 patients with CD to identify oral lesions and the prevalence of gingivitis and periodontitis. The results showed a higher incidence of periodontitis in CD patients compared to the control group (P = 0.01). However, since there were only two cases of periodontitis in the CD group and none in the control group, the calculated OR had a very wide CI (0.38-179.82), leading to its exclusion from the meta-analysis.

Our study aimed to determine the OR of periodontitis within a population diagnosed with IBD, necessitating the inclusion of only those studies that provided a definitive diagnosis of periodontitis. The meta-analysis encompassed 1,292 cases of IBD and 979 control subjects, revealing a significant increase in the risk of periodontitis among patients with IBD compared to the control group. Quantitative analyses revealed that eight of the studies included revealed a significant positive link between periodontitis and IBD [23, 32,33,34,35,36,37,38]. For instance, a case-control study conducted by Grössner et al. in 2006 [24] found that probing pocket depth was significantly higher in the control group compared to IBD patients (P = 0.01). While IBD patients had a greater number of sites with clinical attachment loss of ≥ 4 mm and ≥ 5 mm compared to non-IBD patients, this difference was not statistically significant. In another study by Zervou et al. in 2004 [31], it was reported that the incidence of periodontitis was greater in individuals with CD than in the control group, although this trend was not observed in those with UC. Additionally, Majster et al. in 2024 [39] found no significant difference in the prevalence of periodontitis between IBD patients and the control group.

This systematic review included studies focused on participants aged 18 years and older. Additionally, only studies that offered a clear, clinical, and comprehensive definition of periodontitis were reviewed and analyzed. For instance, a study that employed the World Health Organization questionnaire to define periodontitis was excluded from this analysis [40]. While this exclusion may limit the applicability of the findings, the authors deemed it crucial to define periodontitis using clinical indices as the gold standard.

Previous systematic reviews have explored the association between periodontitis and IBD. Zhang et al. (2021) [41] conducted their investigation utilizing three databases—Web of Science, PubMed, and Embase—and incorporated studies that defined periodontitis through both clinical assessment and self-reporting. In contrast, our review undertook a comprehensive search across seven databases and employed rigorous clinical diagnostic criteria, thereby minimizing the risk of misclassification bias associated with self-reported information. Lorenzo-Pouso AI et al. (2021) [42] incorporated radiographic bone loss as a diagnostic criterion for periodontitis. While bone loss is a relevant marker, our study relied on clinical assessments to ensure diagnostic consistency across included studies. Ayati et al. (2025) [43] included self-reported periodontitis (Bertl et al.’s study [40], which used questionnaires rather than clinical examinations), whereas our analysis exclusively used clinical diagnostic criteria (the gold standard). They did not impose an age restriction (≥ 18 years), leading to the inclusion of pediatric populations (e.g., Koutsochristou et al.’s study [44]). Since periodontitis pathophysiology and IBD manifestations differ between children and adults, this may have introduced heterogeneity. Their review used the JBI checklist for quality assessment, which relies on yes/no/unclear responses, whereas our study employed the Newcastle-Ottawa Scale (NOS), a more quantitative, star-based system extensively validated for observational studies.

In all studies reviewed, the presence of concomitant systemic diseases that could potentially impact the immune response of participants was established as an exclusion criterion. Consequently, the influence of other systemic conditions and diseases was effectively mitigated. Participants in the studies included in this systematic review were exclusively receiving medications pertinent to IBD. This uniformity in treatment regimens across the various studies ensured consistency in drug type and mechanism of action. The careful exclusion of confounding systemic conditions and the homogeneity of treatment protocols helped improve internal validity.

The current study is subject to several limitations. Firstly, there was a lack of consensus regarding the definition of periodontitis across some included studies, which could have introduced variability and potential bias in the results. Moreover, the accuracy of the reported data, including periodontal indices and the prevalence of periodontitis, may be compromised by measurement errors inherent in the methodologies employed. Besides the small sample sizes of the research studies included in this study, it is important to note that smoking is recognized as a significant risk factor for periodontitis. Individuals who smoke not only face an increased risk of developing periodontitis but are also more susceptible to IBD, such as CD. Furthermore, the presence of diversity within the studied populations, including variations in ethnic and geographic genetic backgrounds and differing dietary habits and lifestyles among participants across various studies, represents significant limitations and confounding factors in this research. Nearly all studies either failed to report confounding variables or had significant flaws, making meta-regression unfeasible.

It is important to highlight that all the studies included in this review were of observational type, which means that the review is not able to determine the direction of the relationship or causation. This leaves the question unresolved as to whether IBD contributes to worsening periodontal health, whether periodontitis may worsen IBD, or if other factors make individuals susceptible to both conditions. Additionally, confounding variables like diabetes, oral hygiene habits, and socioeconomic status were not considered in all studies. Moreover, some studies presented data that could not be statistically analyzed. The lack of sufficient data to assess smoking as a significant confounder in these studies prevented the possibility of conducting meta-regression. Although attempts were made to include merely comparable groups, there may still be uncontrolled differences between the IBD and non-IBD groups (such as access to healthcare or antibiotic usage) that could influence oral health. Furthermore, the small sample sizes of the included studies may not represent all patient subgroups, such as those with severe versus mild IBD or those undergoing different treatments. Most of the studies we included had small sample sizes that may introduce bias (e.g., overestimation or underestimation of the association), reduce statistical power, or limit generalizability. Although our sensitivity analysis suggested that no single study disproportionately influenced the pooled OR for periodontitis, the small sample size may still increase the risk of overestimation due to unpublished negative findings or heterogeneity among studies. Readers should interpret the effect size with caution until larger-scale meta-analyses confirm these results. This point underscores the need for larger, high-quality cohorts to improve estimate precision. Future researches are also recommended to focus on younger populations.

Several studies incorporated in this systematic review assessed periodontal indices alongside the prevalence of periodontitis. The analysis of these findings revealed an elevated risk of clinical attachment loss and probing pocket depth in individuals with IBD compared to those without the condition [23, 32, 33, 35, 37].

The oral cavity and intestines, as the terminal regions of the digestive system, exhibit comparable microbial pathogenic mechanisms. Disruptions in the intestinal microbial balance can lead to alterations in bacterial metabolites, releasing numerous pathogenic factors that may compromise host tissues and trigger an inflammatory response [45]. The human gut microbiota predominantly consists of Firmicutes, Bacteroides, Proteobacteria, and Actinobacteria, with Firmicutes and Bacteroides being the most prevalent [46]. The host’s tissues create a nutrient-rich environment that supports the gut microbiota, contributing to the host’s health by producing short-chain fatty acids and essential vitamins, thereby maintaining a state of mutual symbiosis. However, in the context of IBD, alterations in the gut microbiota occur, a phenomenon referred to as “dysbiosis” [47].

According to Schmidt et al.’s research in 2019 [48], oral microorganisms are present in fecal samples, albeit at low concentrations. Oral microorganisms can migrate to the gastrointestinal tract, facilitating the colonization of pathogenic bacteria that disrupt intestinal homeostasis. This disruption may elicit an enhanced immune response from the host, thereby contributing to the pathogenesis of IBD [49]. Furthermore, the oral cavity may be a significant reservoir for enteric pathogens. For instance, the onset of periodontitis is associated with an increase in the populations of Klebsiella and Proteus within the oral cavity. These microorganisms can subsequently migrate to the lower gastrointestinal tract, where they might colonize inappropriately, leading to the release of inflammatory mediators in macrophages, potentially exacerbating intestinal inflammation [50].

Porphyromonas gingivalis (P. gingivalis) and Fusobacterium nucleatum (F. nucleatum), recognized as pathogenic bacteria associated with periodontitis, can impact IBD’s progression. Research conducted by Lee et al. in 2022 [51] demonstrated a significant increase in the presence of P. gingivalis in the feces of patients with CD. This intestinal colonization contributed to the deterioration of the intestinal surface epithelium, disruption of crypt architecture, and infiltration of inflammatory cells, thereby exacerbating the inflammatory symptoms within the gastrointestinal tract. Strauss et al. [52] identified the presence of F. nucleatum in the intestinal mucosa of patients with IBD. Their findings indicate a positive correlation between the invasive potential of various F. nucleatum strains and the severity of the host’s disease status, suggesting that F. nucleatum may serve as a valuable biomarker for gastrointestinal disorders. Beyond the presence of microbes, their metabolites can traverse the gut-mouth axis, contributing to heightened intestinal inflammation [19].

Furthermore, systemic inflammation in individuals with IBD has been shown to influence oral microbiota composition. Research conducted by Docktor et al. [53] revealed an increase in the abundance of Spirochetes and Bacteroides, alongside a reduction in Firmicutes and Fusobacteria, within the oral cavity of patients diagnosed with IBD. Said et al. [54] conducted a study utilizing bacterial 16 S rRNA sequencing and identified a notable increase in the abundance of Prevotella spp. within the salivary microbiota of individuals diagnosed with IBD. In contrast, the study also revealed a significant reduction in the prevalence of Streptococcus species, typically the most abundant genera found in the saliva of healthy individuals.

While our meta-analysis affirms an epidemiological link between periodontitis and IBD, the association may be attributed to overlapping inflammatory pathways and dysbiosis-driven immune responses. Both conditions are characterized by chronic inflammation, which is mediated by pro-inflammatory cytokines such as TNF-α, IL-1β, and IL-6, alongside the aberrant activation of immune cells, including Th17 cells. These factors contribute to tissue damage in both the gastrointestinal and periodontal systems. Moreover, microbial dysbiosis is pivotal in the pathogenesis of IBD, as an altered gut microbiome exacerbates intestinal inflammation. In the case of periodontitis, pathogenic oral bacteria, like P. gingivalis, are key players in the disruption of periodontal integrity. Microbial communities in the oral cavity and gastrointestinal tract may engage in systemic interactions, where translocation of oral pathogens to the gut can worsen the severity of IBD. Conversely, inflammation in the gut may enhance periodontal inflammation through the release of circulating inflammatory mediators. Genetic susceptibility such as caspase activating recruitment domain 15/nucleotide oligomerisation domain 2 (CARD15/NOD2) mutations and environmental risk factors like smoking further link the two conditions, suggesting a bidirectional relationship fueled by shared immunopathological mechanisms [55,56,57].

Another investigation demonstrated that IBD significantly alters bacterial colonization within the oral cavity of affected patients. Notably, the degree of this disruption correlates with the severity of the disease. Furthermore, it was observed that these microbial imbalances tend to resolve gradually following the effective treatment of IBD [58]. Since the current study solely focused on observational research, it is still unclear if treatments for IBDs positively impact the periodontal health of affected patients. Therefore, it is suggested that additional clinical trials and systematic reviews be conducted, with an emphasis on controlling for confounding variables.

Host immune factors play a critical role in the pathogenesis of periodontitis and IBD. Immune cells and inflammatory cytokines can disseminate from localized sites through the bloodstream, potentially leading to inflammatory responses in distant organs. The expression levels of cytokines serve as valuable biomarkers for evaluating the clinical status of patients suffering from periodontitis and IBD. Notably, the IL-17/IL-23 signaling pathway is instrumental in both conditions, as it modulates the innate immune response to various tissues and pathogens [30, 59].

Suggestions for practice and research

Despite the complexities inherent in both IBD and periodontitis, additional research is essential to elucidate the underlying mechanisms of IBD and to refine its treatment strategies. A significant number of patients with IBD may not fully recognize the implications of their condition about the onset and progression of periodontal disease. Consequently, physicians, as integral healthcare team members, have a pivotal role in educating patients about these connections and facilitating referrals for regular periodontal assessments and care as part of IBD management. During clinical follow-ups, patients should be advised to undergo comprehensive periodontal screenings at least annually, adopt preventive oral hygiene practices and seek prompt treatment for gingival inflammation to reduce systemic inflammatory burden.

Future research should address current limitations, including heterogeneity in periodontitis definitions (e.g., standardization via the 2017 World Workshop criteria) and confounding factors (smoking, diabetes, and medication effects). Large-scale, longitudinal studies with adjusted confounders are needed to establish causality and elucidate underlying mechanisms, such as the role of oral-gut microbial translocation or shared immunogenetic pathways. Additionally, interventional studies assessing the impact of periodontal therapy on IBD activity could inform clinical guidelines and improve patient outcomes.