TCR-Pred - Prediction of epitopes and MHC type for CDR3 TCR sequences

This resource is about prediction of epitopes and HLA types for alpha and beta chains of CDR3 T-cell receptors (TCR) amino acid sequences based on classification structure-activity relationships (SAR) models. SAR models were created based on the data from VDJdb, McPAS-TCR and IEDB databases using Bayesian-like approach and atom centric substructural Multilevel Neighborhoods of Atoms (MNA) descriptors. During the prediction the letter text representation of CDR3 TCR amino acid sequences are transformed to MOL V3000 format of structural formula representation.

Prediction is based on special version of PASS (Prediction of Activity Spectra for Substances) technology (http://www.way2drug.com/PASSonline) which was modified for generation up to 15 levels of MNA descriptors. After creation of SAR models with different levels of MNA descriptors we selected level of MNA descriptors leading to the most accurate SAR models. The SAR models for prediction of epitopes were created based on 9st (for alpha chain of CDR3 TCR) or 11st (for beta chain of CDR3 TCR) level of MNA descriptors. The SAR models for prediction of MHC types were created based on 11st (for beta chain of CDR3 TCR) or 8st (for alpha chain of CDR3 TCR) level of MNA descriptors. Lists of predicted epitopes and MHC types with accuracy of prediction are available at description of Training Sets.

Please cite us: Smirnov AS, Rudik AV, Filimonov DA, Lagunin AA. TCR-Pred: A new web-application for prediction of epitope and MHC specificity for CDR3 TCR sequences using molecular fragment descriptors. Immunology. 2023, 169(4):447-453. doi: 10.1111/imm.13641 https://onlinelibrary.wiley.com/doi/10.1111/imm.13641