Details of UCE's Cross-Species Representation
jure · x · 2026-07-14
This post further explains UCE's technical approach: the model connects molecular and cellular scales, no longer treating genes merely as columns in an expression matrix.
In UCE, genes are represented by the proteins they encode and embedded using ESM. This allows the model to generalize to novel species unseen during training, covering data across an evolutionary scale of 100 million years.
Related event: UCE Single-Cell Foundation Model Published in Nature(5 posts)→
More from Research
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- World Model Assigns Weight and Friction to Objects, Then Simulates with a Physics Engine — jonstephens85 · 2026-10-27
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27