Gene-Family Encoding Boosts Domain-Adapted Single-Cell Language Models
Haoran Ma, Chang Xu, Shamaine Wei Ting Ho, Joseph Jonathan Zhao, Yunqiang Chu, Angie Lay Keng Tan, Raghav Sundar, Patrick B. O. Tan
This paper introduces gene-family encoding to improve domain-adapted single-cell language models, showing better performance in embedding-based and generative tasks on gastric-cancer data.
bioRxiv · Sep 16, 2025 · PreprintBioinformatics