Fresh negative examples per batch boost TCR-pHLA binding prediction AUROC by 5.22%
bravo_abad · x · 2026-10-29
Zhu et al. tackle a core problem in T-cell receptor–pHLA binding prediction: an untested pair isn't necessarily a nonbinder, and reshuffling known pairs into a fixed negative set teaches the model quirks of that selection. Their fix: generate fresh negative pairs for every training batch, excluding known binders while preserving each fragment's frequency in the positive data, then fine-tune ESM2 protein language models. On an external experimental benchmark this improved AUROC by 5.22% relative to the strongest fixed-negative baseline for the 35M-parameter model, published in Nature Computational Science.
More from Research
- Axiom's agents are formalizing CFSG in 43M lines of code, possibly the largest open-source project ever — SucceededMind · 2026-10-29
- MIT's 'AI telepathy' KV-cache transfer beats text by 40% accuracy at 4x speed, across model vendors — LucaAmb · 2026-10-29
- Coding agents fail to reimplement models from papers, sparking doubts about how papers are written — charlieharris01 · 2026-10-29
- Information-theoretic scaling laws for diffusion models predict when memorization begins — davidad · 2026-10-29
- Amandine Flachs launches side project orbench — IgorCarron · 2026-10-29
- Critic says closed weights hide personas from users, but that doesn't prove models aren't general-purpose actors — aiamblichus · 2026-10-29