Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines.
Ahlmann-Eltze, Huber, Anders: deep models don't beat simple linear baselines for perturbation prediction.
- Authors / organisation
- Ahlmann-Eltze C, Huber W, Anders S.
- Venue
- Nature methods
- Date
- 2025-08-04
- Type
- paper
- Year
- 2025
- DOI
- 10.1038/s41592-025-02772-6
- PMID
- 40759747
- Verification
- Europe PMC
Cited by GlioState topics
- Mesenchymal-like (MES-like) glioblastoma cell state (Cell states)
- AI virtual cell models (Models)
- Perturbation data for glioblastoma: radiotherapy, drugs and CRISPR (Perturbations)
- Therapy-induced plasticity in glioblastoma (Transitions)
- Validating virtual cells: benchmarks, baselines and critiques (Validation)
- Do Neftel-style GBM state calls replicate across platforms? A GlioState re-analysis (draft) (Validation)
Entities mentioned
- Simple linear baseline Methods & concepts
- Perturbation response prediction Methods & concepts