Benchmarking solutions to the T-cell receptor epitope prediction problem: IMMREP22 workshop report

• DOI: 10.1016/j.immuno.2023.100024
• Publication Date: 2023-02-03T07:35:04Z
• Authors: Pieter Meysman; Justin Barton; Barbara Bravi; Liel Cohen-Lavi; Vadim Karnaukhov; Elias Lilleskov; Alessandro Montem...; Morten Nielsen; Thierry Mora; Paul Pereira; Anna Postovskaya; María Rodríguez M...; Jorge Fernandez-d...; Alexandra Vujkovic; Aleksandra M. Wal...; Anna Weber; Rose Yin; Anne Eugster; Virag Sharma
• Keywords: Benchmark (surveying); Benchmarking

Many different solutions to predicting the cognate epitope target of a T-cell receptor (TCR) have been proposed. However several questions on advantages and disadvantages these approaches remain unresolved, as most methods only evaluated within context their initial publications data sets. Here, we report findings first public TCR-epitope prediction benchmark performed 23 models in ImmRep 2022 specificity workshop. This revealed that use paired-chain alpha-beta, well CDR1/2 or V/J information, when available, improves classification obtained with CDR3 data, independent underlying approach. In addition, found straight-forward distance-based can achieve respectable performance compared more complex machine-learning models. Finally, highlight need for truly follow-up provide recommendations design such next benchmark.