Large-scale template-based structural modeling of T-cell receptors with known antigen specificity reveals complementarity features.

After repacking of modeled structures, which led to structural fitting of contacting residues and displacement of their side-chain groups, the correlations of steric descriptors became weaker, while the correlations of hydrophobic descriptors still remained highly significant (p-value = 6x10 -5 ). 3.4 Identifying residue binding features in CDR3: peptide interface using machine learning At the final stage we performed RFE (Recursive Feature Elimination) analysis and applied the Random Forest (RF) algorithm to make predictive models for the per-residue contacting energy values for amino acid residue pairs in CDR3 loops and peptides.

also in 132,142 other papers