BenchRep-T: A Systematic Evaluation of T-Cell Repertoire-Based Disease Diagnostics

BenchRep-T shows that, across four T-cell receptor repertoire diagnostic tasks, no single computational method dominates, while V/J-gene and short-motif tree models can approach top classification performance.

• Why it matters: Directly comparing TCR repertoire disease diagnostics is difficult because published studies use different cohorts, preprocessing pipelines, and evaluation metrics.
• What they did: BenchRep-T standardizes multiple public TCR repertoire datasets and evaluates nine computational approaches across disease classification, sampling-depth scaling, recovery of known antigen-specific driver sequences, and demographic confounding sensitivity.
• The result: In controlled evaluations, simple baselines are competitive and tree-based models using V- and J-gene usage plus short sequence motifs can approach the performance of more complex methods, enabling rigorous, reproducible benchmarking.

AI summary of the abstract · may contain errors · not medical advice · read the paper.