Yuanhua Huang
banner
yuanhuahuang.bsky.social
Yuanhua Huang
@yuanhuahuang.bsky.social
At the junction between data science and cell biomedicine
It's a pity to miss #ISMBECCB2025 & this exciting event. Just want to share a bit on our recent try with genomic language models for personal gene expression prediction.
bsky.app/profile/yuan...
New preprint: A big headache for sequence models is to predict cross-person variability of RNA levels from DNA in zero-shot genes. Our gLM2X-Tower shows the problem remains unsolved, incl Evo2 & AlphaGenome. However, few-shot setting is promising & can be a new focus.
www.biorxiv.org/content/10.1...
Assessing large-scale genomic language models in predicting personal gene expression: promises and limitations
Large-scale genomic language models (gLMs) hold promise for modeling gene regulation, yet their ability to predict personal gene expression remains largely unexplored. We developed a framework, gLM2X-...
www.biorxiv.org
July 21, 2025 at 7:44 AM
@ShuminLi led the development of the gLM2X-Tower framework and all benchmarking analysis, with computing support from @RuibangLuo!
This work was also used for our Genomic AI hackathon, supported by @hkusbms.bsky.social and @CPOS!
bsky.app/profile/yuan...
Our full-day mini AI hackathon on genomic language models starts soon: 8.30am-8.30pm! With a whole team, we aim to test how gLMs work for two tasks:
1. SNP2GEX: personal variants to gene expression (unseen individuals & genes);
2. Seq2CellxTF: short regulatory to TF binding (unseen cell & TFs).
July 21, 2025 at 2:35 AM