Fig. 5: Evaluating antibody breadth on synthetic Omicron lineages.
a, Example of a synthetic lineage tree of sequences generated containing mutations unseen in major Omicron variants, with heat map indicating the deep learning predictions of binding or escape for individual antibodies. VOCs, variants of concern. b, Total mean predicted breadth of individual antibodies and combinations on synthetic lineages generated from 2022 mutational probabilities. c, The fraction (%) of sequences bound by individual antibodies at different ED from BA.1. d, Phylogenetic tree of ZCB11 escape variants containing the 20 highest-scoring mutations, with antibody recapture indicated. e, Sequence logos show the mutations in the top 25 positions with greatest KL divergence in ZCB11 escape variants at ED6, and number of sequences re-captured by Brii-198, ADG20 and A23-58.1. (Higher recapture indicates a more complementary antibody). f, The top 50 predicted mutations ranked by their escape scores (Methods) from the generated synthetic lineages, with new mutations seen in the BA.2.86 variant highlighted.
