Back

Mutational basis of serum cross-neutralization profiles elicited by infection or vaccination with SARS-CoV-2 variants

Wagh, K.; Shen, X.; Theiler, J.; Girard, B.; Marshall, J.-C.; Montefiori, D.; Korber, B.

2023-08-14 immunology
10.1101/2023.08.13.553144 bioRxiv
Show abstract

A series of SARS-CoV-2 variants emerged during the pandemic under selection for neutralization resistance. Convalescent and vaccinated sera show consistently different cross-neutralization profiles depending on infecting or vaccine variants. To understand the basis of this heterogeneity, we modeled serum cross-neutralization titers for 165 sera after infection or vaccination with historically prominent lineages tested against 18 variant pseudoviruses. Cross-neutralization profiles were well captured by models incorporating autologous neutralizing titers and combinations of specific shared and differing mutations between the infecting/vaccine variants and pseudoviruses. Infecting/vaccine variant-specific models identified mutations that significantly impacted cross-neutralization and quantified their relative contributions. Unified models that explained cross-neutralization profiles across all infecting and vaccine variants provided accurate predictions of holdout neutralization data comprising untested variants as infecting or vaccine variants, and as test pseudoviruses. Finally, comparative modeling of 2-dose versus 3-dose mRNA-1273 vaccine data revealed that the third dose overcame key resistance mutations to improve neutralization breadth. HIGHLIGHTSO_LIModeled SARS-CoV-2 cross-neutralization using mutations at key sites C_LIO_LIIdentified resistance mutations and quantified relative impact C_LIO_LIAccurately predicted holdout variant and convalescent/vaccine sera neutralization C_LIO_LIShowed that the third dose of mRNA-1273 vaccination overcomes resistance mutations C_LI

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.