Abstract
Abstract The global emergence of SARS-CoV-2 has provided a wealth of immunological data, including more than 600 solved structures of SARS-CoV-2 Spike antibodies (Abs) available in the Protein Data Bank. We reasoned that this data could be used to generate a computational model of the serum Ab repertoire. We extracted a set of 275 non-redundant Abs, and clustered them using an unsupervised clustering method based on their contact positions with Spike. We identified 9 unique Ab clusters that together summarize the major binding modes of Abs to the Spike protein. To characterize each cluster, we mined 65 published papers extracting detailed information on each Ab. We then demonstrated Spike Ab contacts could be predicted based on the Ab’s CDRs sequence. Next, we developed a computational method for predicting immune escape. We extracted the set of all Spike mutations in SARS-CoV-2 sequences from GISAID. To assess the effects of single mutations on immune escape, we computed the ΔΔG for each Ab-mutation pair and constructed an aggregated escape score for each mutation. Our predicted escape scores were positively correlated with published experimental measurements of Ab escape. We used our escape score to rank novel viral variants and are dynamically ranking strains and novel variants of interest as they appear. To further validate its predictive power, we chose a set of predicted Ab-escaping mutations and are testing their escape from Abs in human sera samples. Taken together, our analysis demonstrates the feasibility of generating a computational model of the serum Ab repertoire, which may be used for the rapid characterization of novel Abs, and to computationally estimate the effects of mutations on immune escape from Ab recognition. Supposed by the Israel Science Foundation (ISF) grant no. 2683/21, and the NIH\NIAID CEIRR under contract 75N9302100016, St Jude Children’s Research Hospital.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have