New AI-powered platform could help scientists and health officials catch the next COVID-19 variant before it spreads, offering the world a crucial head start in the fight against future pandemics.
Study: In silico genomic surveillance by CoVerage predicts and characterizes SARS-CoV-2 variants of interest. Image credit: peterschreiber.media/Shutterstock.com
Researchers at the Helmholtz Centre for Infection Research and the German Center for Infection Research developed a web-based platform to identify and characterize concerning variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) early in their development. The study is published in Nature Communications.
Background
SARS-CoV-2, the causative pathogen of the coronavirus disease 2019 (COVID-19) pandemic, is a single-stranded, positive-sense RNA virus with a high capacity of acquiring mutations during its evolution. These mutations can potentially increase the transmissibility, pathogenicity, or immune escape capacity of the virus, leading to emergence of more infectious or more harmful variants, designated as Variants of Concern (VOC) or Variants of Interest (VOI) by the World Health Organization (WHO).
A high immune escape capacity allows SARS-CoV-2 to evade anti-viral immunity developed through previous infection or vaccination. This highlights the need to frequently upgrade COVID-19 vaccines to maintain their effectiveness against circulating variants.
Large-scale viral genomic surveillance programs have been implemented in several countries worldwide to continuously monitor SARS-CoV-2 evolution and adaptation and timely identification of new VOCs. This has led to the generation of a vast amount of viral genome sequencing data in the GISAID database. Although the GISAID database has immensely helped researchers and public health officials characterize viral evolution, methods remain needed to continuously interpret these sequences and promptly ensure the continued efficacy of vaccines.
In the current study, researchers developed an online analysis method, the CoVerage system, for the genomic surveillance of SARS-CoV-2.
The CoVerage system
The CoVerage system analyzes SARS-CoV-2 genomic sequence data from the GISAID database, which contains more than 16.5 million sequences. The system continuously predicts and characterizes emerging potential VOIs by country of origin for strain dynamics and antigenic changes.
The system includes a suite of statistical and bioinformatic methods, including Fisher’s exact test and correction for multiple comparisons, that compares the mutations occurring in the spike protein on the surface of different viral strains in a given month. Viral strains with significantly higher mutations than the average are predicted to have higher transmissibility or immune escape capacity. They are subsequently displayed on the CoVerage platform in special graphics called “heatmaps” so that users can see when and where significant changes in the virus are occurring.
System validation
The researchers tested the reliability of the CoVerage system by analyzing genome sequence data of known VOCs, including the Omicron variant of SARS-CoV-2. They observed that the system can identify these sequences as VOCs on average 79 days before the WHO designation.
The system utilized a method that scores amino acid changes based on a viral immune escape capacity to identify SARS-CoV-2 variants with antigenic alterations. These antigenic alteration scores are calculated using a matrix that weighs mutations across the entire spike protein, not just at previously known antigenic sites. They are benchmarked against experimental neutralization data for validation.
In the heatmaps, these antigenic alteration scores increased in a clear order, firstly displaying variants that are only being monitored, followed by VOIs, and finally, most strongly, the VOCs, which are considered particularly harmful.
Study significance
The study describes the development and validation of a genomic surveillance platform, CoVerage, that continuously monitors incorporated SARS-CoV-2 genome sequence data to identify and characterize potential VOIs from circulating viral strains in a timely manner. It also suggests their degree of antigenic alterations and alleles of spike protein with specific amino acid changes that may provide a selective advantage.
The CoVerage system includes three novel methods: one method detects potential VOIs with higher transmissibility; a second method analyzes the dynamics of amino acid changes throughout the major surface spike proteins to pinpoint those that may confer a selective advantage; and a third method that scores the degree of antigenic alteration of each variant using a unidirectional immune escape matrix.
The systemic assessment of CoVerage indicates that the system can identify 88% of the VOIs and VOCs designated by the WHO, with a precision of 79% and recall of 72%, more than two months before their official WHO designation. No VOCs were missing, and most of the missed lineages were lower public health relevance (Variants Under Monitoring).
The predictions made by CoVerage depend on the extent and quality of ongoing viral genomic surveillance programs for individual countries. The analysis is done country-wise and may also be affected by population genetic effects when case numbers are low. Any reduction in genomic surveillance can thus affect its predictive capacity.
Several other web-based platforms, including NextStrain, CoVariants, CovidCG, EVEscape, and SpikePro, monitor SARS-CoV-2 variants and characterize their mutagenic frequencies. However, none of these platforms continuously score all circulating variants for potential advantage and antigenic change in real time. They also don’t provide benchmarking against experimental antigenicity data as CoVerage does.
Furthermore, the CoVerage system combines GISAID data with links to alternative web-based resources. It offers reproducible, open-access analytics for additional information on selected variants, providing a comprehensive resource for viral genomic surveillance.
Download your PDF copy now!