Dynamic Positive-Unlabeled graph learning separates biology from sampling bias in virus-host prediction
bravo_abad · x · 2026-09-22
Missing links in scientific datasets are often silently treated as negatives. For mammal-virus associations, sampling effort is highly uneven across species, viral families and regions. Pignalberi et al. propose a Dynamic Positive-Unlabeled framework that jointly trains a graph neural network estimating biological plausibility of associations and a separate propensity model estimating observation likelihood, disentangling true non-interactions from sampling gaps in virus-host prediction.
More from Research
- PIR lie-detector reads LLM internal states to expose hidden knowledge at 0.70-0.87 accuracy — Hiskias Dingeto · 2026-09-22
- TAPe+ML: sub-100K-parameter vision system hits 84.7 mAP50 on COCO detection — Comexp · 2026-09-22
- RoboDawn: Tsinghua & Tencent Hunyuan drive robots with frozen VLM, 73.6% one-shot success — _akhaliq · 2026-09-22
- PragmaBot: robots learn online from real-world failures without retraining, RA-L/IROS 2026 — ChongZzZhang · 2026-09-22
- Iterative sensitivity probing: how bitsandbytes2 finds each layer's compression limit — Tim_Dettmers · 2026-09-22
- Tim Dettmers releases runtime dynamic compression framework, hits 1.5-2.0 bit at high quality — Tim_Dettmers · 2026-09-22