Preprint: data leakage yields AUC near 0.8 from pure noise in proteomics
bttyeo · x · 2026-09-11
A new preprint quantifies a common flaw in proteomics biomarker studies—data leakage: with NO true signal, some leaked runs still reached AUCs near 0.8.
The team (JakeVogel, Caitlin Finney, Artur Shvetcov et al.) notes leakage is widespread; even simple mistakes produce AUCs >0.7 from pure noise while worsening generalization. They built tools to help researchers detect and avoid leakage and are pushing for trustworthy reporting standards.
Related event: Preprint Quantifies Data Leakage: Near-0.8 AUC With No Signal(2 posts)→
More from Research
- Rachel Thomas points to classic paper on articulating the value of not building tech — math_rachel · 2026-10-28
- Meta's Omnilingual MT covers 1,600 languages, 1B–8B models rival 70B LLMs — kalyan_kpl · 2026-10-28
- Supercharging LLM classification with logprobs: soft classifiers you can threshold-tune — JnBrymn · 2026-10-28
- cryoET Data Portal releases three datasets benchmarking crossed laser phase plate — ZhongingAlong · 2026-10-28
- Blind Robot Hand Spins Baoding Balls 225 Times With Only Touch — Scobleizer · 2026-10-28
- Rollback: World Models Rewind Robot Decisions to Explore Counterfactual Actions — agibot-world · 2026-10-28