Figure 2

Visualization of the input dataset. We visualized the input dataset before feature extraction and ML analysis as follows: (A) Numbers of neutralizing (n = 554) and nonneutralizing (n = 554) interactions based on target epitope domains are shown in a bar plot. EDI is envelope domain I. EDII is envelope domain II. EDIII is envelope domain III. EDI-EDII is interdomain. EDE is the envelope dimer epitope domain. (B) Diversity of CDR-H3 antibody sequences based on target domains including EDI, EDII, EDIII, interdomain, and EDE by t-SNE (n = 306; perplexity = 30, learning rate = 100), (C) Diversity of epitope sequences based on target domains including EDI, EDII, EDIII, interdomain, and EDE by t-SNE (n = 609; perplexity = 30, learning rate = 100), and (D) Distribution of IC50 values of each antibody–antigen interaction by scatter plot (n = 1,108; cut-off value for neutralizing class ≤ 10 μg/ml).