An explainable language model for antibody specificity prediction using curated influenza hemagglutinin antibodies.

Despite decades of antibody research, it remains challenging to predict the specificity of an antibody solely based on its sequence. Two major obstacles are the lack of appropriate models and the inaccessibility of datasets for model training. In this study, we curated >5,000 influenza hemagglutinin (HA) antibodies by mining research publications and patents, which revealed many distinct sequence features between antibodies to HA head and stem domains. We then leveraged this dataset to develop a lightweight memory B cell language model (mBLM) for sequence-based antibody specificity prediction. Model explainability analysis showed that mBLM could identify key sequence features of HA stem antibodies. Additionally, by applying mBLM to HA antibodies with unknown epitopes, we discovered and experimentally validated many HA stem antibodies. Overall, this study not only advances our molecular understanding of the antibody response to the influenza virus but also provides a valuable resource for applying deep learning to antibody research.

Leveraging vaccination-induced protective antibodies to define conserved epitopes on influenza N2 neuraminidase.

There is growing appreciation for neuraminidase (NA) as an influenza vaccine target; however, its antigenicity remains poorly characterized. In this study, we isolated three broadly reactive N2 antibodies from the plasmablasts of a single vaccinee, including one that cross-reacts with NAs from seasonal H3N2 strains spanning five decades. Although these three antibodies have diverse germline usages, they recognize similar epitopes that are distant from the NA active site and instead involve the highly conserved underside of NA head domain. We also showed that all three antibodies confer prophylactic and therapeutic protection in vivo, due to both Fc effector functions and NA inhibition through steric hindrance. Additionally, the contribution of Fc effector functions to protection in vivo inversely correlates with viral growth inhibition activity in vitro. Overall, our findings advance the understanding of NA antibody response and provide important insights into the development of a broadly protective influenza vaccine.

Epistasis mediates the evolution of the receptor binding mode in recent human H3N2 hemagglutinin.

The receptor-binding site of influenza A virus hemagglutinin partially overlaps with major antigenic sites and constantly evolves. In this study, we observe that mutations G186D and D190N in the hemagglutinin receptor-binding site have coevolved in two recent human H3N2 clades. X-ray crystallography results show that these mutations coordinately drive the evolution of the hemagglutinin receptor binding mode. Epistasis between G186D and D190N is further demonstrated by glycan binding and thermostability analyses. Immunization and neutralization experiments using mouse and human samples indicate that the evolution of receptor binding mode is accompanied by a change in antigenicity. Besides, combinatorial mutagenesis reveals that G186D and D190N, along with other natural mutations in recent H3N2 strains, alter the compatibility with a common egg-adaptive mutation in seasonal influenza vaccines. Overall, our findings elucidate the role of epistasis in shaping the recent evolution of human H3N2 hemagglutinin and substantiate the high evolvability of its receptor-binding mode.

An explainable language model for antibody specificity prediction using curated influenza hemagglutinin antibodies.

Despite decades of antibody research, it remains challenging to predict the specificity of an antibody solely based on its sequence. Two major obstacles are the lack of appropriate models and inaccessibility of datasets for model training. In this study, we curated a dataset of >5,000 influenza hemagglutinin (HA) antibodies by mining research publications and patents, which revealed many distinct sequence features between antibodies to HA head and stem domains. We then leveraged this dataset to develop a lightweight memory B cell language model (mBLM) for sequence-based antibody specificity prediction. Model explainability analysis showed that mBLM captured key sequence motifs of HA stem antibodies. Additionally, by applying mBLM to HA antibodies with unknown epitopes, we discovered and experimentally validated many HA stem antibodies. Overall, this study not only advances our molecular understanding of antibody response to influenza virus, but also provides an invaluable resource for applying deep learning to antibody research.

Prevalence and mechanisms of evolutionary contingency in human influenza H3N2 neuraminidase.

Neuraminidase (NA) of human influenza H3N2 virus has evolved rapidly and been accumulating mutations for more than half-century. However, biophysical constraints that govern the evolutionary trajectories of NA remain largely elusive. Here, we show that among 70 natural mutations that are present in the NA of a recent human H3N2 strain, >10% are deleterious for an ancestral strain. By mapping the permissive mutations using combinatorial mutagenesis and next-generation sequencing, an extensive epistatic network is revealed. Biophysical and structural analyses further demonstrate that certain epistatic interactions can be explained by non-additive stability effect, which in turn modulates membrane trafficking and enzymatic activity of NA. Additionally, our results suggest that other biophysical mechanisms also contribute to epistasis in NA evolution. Overall, these findings not only provide mechanistic insights into the evolution of human influenza NA and elucidate its sequence-structure-function relationship, but also have important implications for the development of next-generation influenza vaccines.