MoESurv: a zero-sample and transferable survival prediction framework for rare cancers using mixture of experts
Shuping Fang, Yuhang Wang, Mengyan Zhou, Yanting Shao, Zhenghao Jiang, Yitong Guo, Tao Jiang, Feng Wang, Jingya Fang, Hong Tian · 2026
Read MoESurv: a zero-sample and transferable survival prediction framework for rare cancers using mixture of experts by Shuping Fang, Yuhang Wang, Mengyan Zhou, Yanting Shao, Zhenghao Jiang, Yitong Guo, Tao Jiang, Feng Wang, Jingya Fang, Hong Tian (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13472725.txt.
Description
Fang and colleagues built MoESurv in 2026, a computer model that predicts survival for rare cancers despite having no training cases of those cancers. Most deep learning models need large datasets, which rare cancers cannot supply. MoESurv instead learns shared patterns from many cancer types at once, using a mixture-of-experts design where separate sub-models handle common and cancer-specific signals.
Tested on seven rare cancer types from a large public atlas, the model beat the best rival method by about four points on a standard ranking score. It also held up in outside cohorts, including a Chinese brain tumor dataset and a rare subtype of glioblastoma, a fast-growing brain cancer. The model split patients into high- and low-risk groups and flagged genes tied to survival, which the authors say adds both clinical and biological value.
Catalog record
| CREATOR | Shuping Fang, Yuhang Wang, Mengyan Zhou, Yanting Shao, Zhenghao Jiang, Yitong Guo, Tao Jiang, Feng Wang, Jingya Fang, Hong Tian |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |