MRFlow: Mechanism-Structured Residual Flow Matching for Virtual Cell Perturbation Prediction
Preprint (not peer-reviewed): Yang and Su (Tianjin University) propose MRFlow, which treats perturbation prediction as latent conditional transport with flow matching; the authors report the highest composite score across eight gene and drug perturbation benchmarks (self-reported, single run).
- Authors / organisation
- Yang H, Su R.
- Venue
- arXiv (preprint, not peer-reviewed)
- Date
- 2026-10
- Type
- preprint
- Year
- 2026
- Source
- https://arxiv.org/abs/2610.13966
- Verification
- arXiv abs page 2610.13966 (title, authors Haojie Yang & Ran Su, abstract, full text read 2026-10-19); exact submission date not retrieved (ID = Oct 2026); no peer-reviewed version found
Cited by GlioState topics
Not yet cited in any research topic page.
Entities mentioned
- Perturbation response prediction Methods & concepts