CGMap: distance-resolved graph learning for cancer gene screening lands in Nature Machine Intelligence
bravo_abad · x · 2026-11-18
Published in Nature Machine Intelligence, CGMap tackles why graph neural networks "blur" at long range: each layer averages a gene with its neighbors, so after a few layers all genes look alike. That hurts cancer gene search — only 800 of 12,000 genes are known, and useful clues sit several hops away.
CGMap keeps distances separate: neighbors one hop away feed one channel, two hops another, up to ten, and the model learns how much to trust each distance, like an equalizer with one slider per hop. The highest-trust distance turned out to be four hops, beyond direct neighbors. Its AUC stays within half a percent whether looking 1 or 10 steps out.
More from Research
- FrontierCS 2: benchmark makes AI agents beat published CS research blind — arena · 2026-11-18
- Isomorphic Labs raising $5B led by Greenoaks and Sequoia as first cancer target nears trials — shiringhaffary · 2026-11-18
- Liquid AI's Maxime Labonne on post-training: SFT, RL, MOPD and regression — ziv_ravid · 2026-11-18
- Self-inoculation cuts RL reward hacking from 82% to 20% across 11 environments — repligate · 2026-11-18
- GPU QR competition audit: winning kernel ran 107x faster yet failed on new inputs — kfountou · 2026-11-18
- GPT-6.2 Astra tops Mercor's OpenWorldEngineer benchmark at 67.2% — morqon · 2026-11-18