Patients with COVID-19 show varying severity of the disease ranging from asymptomatic to requiring intensive care. Although SARS-CoV-2-specific monoclonal antibodies have been identified, we still lack an understanding of the overall landscape of B-cell receptor (BCR) repertoires in patients with COVID-19. We use high-throughput sequencing of bulk and plasma B-cells collected over multiple time points during infection to characterize signatures of the B-cell response to SARS-CoV-2 in 19 patients. Using principled statistical approaches, we can associate differential features of BCRs with different disease severity. We identify 38 significantly expanded clonal lineages shared among patients as candidates for responses specific to SARS-CoV-2. Using single-cell sequencing, we verify the reactivity of BCRs shared among individuals to SARS-CoV-2 epitopes. Moreover, we identify the natural emergence of a BCR with cross-reactivity to SARS-CoV-1 and SARS-CoV-2 in some patients. Our results provide important insights for the development of rational therapies and vaccines against COVID-19.

Proteins play a central role in biology from immune recognition to brain activity. While major advances in machine learning have improved our ability to predict protein structure from sequence, determining protein function from structure remains a major challenge. Here, we introduce Holographic Convolutional Neural Network (H-CNN) for proteins, which is a physically motivated machine learning approach to model amino acid preferences in protein structures. H-CNN reflects physical interactions in a protein structure and recapitulates the functional information stored in evolutionary data. H-CNN accurately predicts the impact of mutations on protein function, including stability and binding of protein complexes. Our interpretable computational model for protein structure-function maps could guide design of novel proteins with desired function.

COVID-19 patients show varying severity of the disease ranging from asymptomatic to requiring intensive care. Although a number of monoclonal antibodies against SARS-CoV-2 have been identified, we still lack an understanding of the overall landscape of B-cell receptor (BCR) repertoires in COVID-19 patients. Here, we used high-throughput sequencing of BCR repertoires collected over multiple time points during an infection to characterize statistical and dynamical signatures of the B-cell response to SARS-CoV-2 in 19 patients with different disease severities. Based on principled statistical approaches, we determined differential sequence features of BCRs associated with different disease severity. We identified 34 significantly expanded rare clonal lineages shared among patients as candidates for a specific response to SARS-CoV-2. Moreover, we identified natural emergence of a BCR with cross-reactivity to SARS-CoV and SARS-CoV-2 in a number of patients. Overall, our results provide important insights for development of rational therapies and vaccines against COVID-19.

Proteins play a central role in biology from immune recognition to brain activity. While major advances in machine learning have improved our ability to predict protein structure from sequence, determining protein function from structure remains a major challenge. Here, we introduce Holographic Convolutional Neural Network (H-CNN) for proteins, which is a physically motivated machine learning approach to model amino acid preferences in protein structures. H-CNN reflects physical interactions in a protein structure and recapitulates the functional information stored in evolutionary data. H-CNN accurately predicts the impact of mutations on protein function, including stability and binding of protein complexes. Our interpretable computational model for protein structure-function maps could guide design of novel proteins with desired function.

No abstract