Simple linear baseline
Also known as: linear baseline · baseline gap
Simple linear models used as reference points for perturbation prediction. Ahlmann-Eltze et al. report that deep-learning gene perturbation prediction does not yet outperform them, and later preprints report that the baseline gap persists. VC0048 VC0054 · preprint VC0053 · preprint
Review pending Version 0.1.0 · status: draft · last expert review: not yet reviewed
Relations
| Subject | Relation | Object | Confidence | Evidence (role) |
|---|---|---|---|---|
| Perturbation response prediction | benchmarked on | Simple linear baseline | high | VC0048 benchmark VC0049 benchmark Deep models do not yet reliably outperform simple linear baselines. |
| X-Cell | benchmarked on | Simple linear baseline | low | VC0033 · preprint supports VC0048 contradicts X-Cell (preprint, not peer reviewed) claims to surpass linear baselines, whereas the Nature Methods benchmark found earlier deep models did not; VC0048 does not test X-Cell, so the “contradicts” role is a tension, not a refutation. Needs review. |
Appears in research topics
- Mesenchymal-like (MES-like) glioblastoma cell state (Cell states)
- AI virtual cell models (Models)
- Validating virtual cells: benchmarks, baselines and critiques (Validation)
Appears in library items (4)
Items whose title, tags or one-line summary match this entity (derived from the data), plus the items cited as evidence for the description.
- VC0054 · preprint An Empirical Comparison of Virtual Cell Models: Perturbation Prediction, Representation, and the Baseline Gap — Denvis O., Research Square, 2026
- VC0110 Causal Models Need Causal Data - Xaira's X-Cell (podcast) — Latent Space, Latent Space podcast, 2026
- VC0053 · preprint Signal, Bounds, and Baselines: Principles for Evaluating Virtual Cell Perturbation Models — Vollenweider M, Bühlmann P., bioRxiv, 2026
- VC0048 Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines. — Ahlmann-Eltze C, Huber W, Anders S., Nature methods, 2025
For research and education only. Not medical advice and not a recommendation of any diagnosis, treatment or product. Patients should consult their own physicians.