Figure 1.

The rationale behind the biochemical approach implementation. (A) Biochemical properties from AAIndex are summarized using dimensionality reduction techniques. The eigenvectors are utilized to convert peptide sequences into biochemical profiles. Next, similar peptides are screened in the immunopeptidomics database. Each pairwise comparison produces a distance-based metric called relatedness score (RdS), which is used for peptide-centered predictions. The relatedness calculation can optionally utilize a position-specific weighted vector, for TCR-centered predictions. Higher values are related with strong penalties (e.g., positions in dark blue). (B) A comprehensive immunopeptidomics database was created by leveraging several public datasets. (C, D) Quantitative description of HLA alleles in the CrossDome database. Image created with BioRender.com.