Research map / Probabilistic and scientific modeling / Molecules and proteins
Beyond Independent Genes: Learning Module-Inductive Representations for Single-Cell Gene Perturbation Prediction
Jiafa Ruan, Ruijie Quan, Xu Liyang, Zongxin Yang, Yi Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use scBIG module-inductive model instead of CellFlow | scBIG · Norman additive split | rho-delta: 0.7892 -> 0.8496 | Improved |
| use scBIG instead of CellFlow on unseen single genes | scBIG · Norman holdout, unseen single gene | rho-delta D: 0.4736 -> 0.6537 | Improved |
| use scBIG instead of Linear-scGPT on unseen double genes | scBIG · Norman holdout, unseen double gene | rho-delta D: 0.7230 -> 0.8719 | Improved |
| use scBIG instead of CellFlow on drug perturbation | scBIG · ComboSciPlex drug perturbation | rho-delta: 0.8276 -> 0.8671 | Improved |
| add GRC clustering instead of balanced K-means (ablation) | scBIG (GCAE fixed) · Norman additive split | rho-delta: 0.8358 -> 0.8496 | Improved |
| add GCAE encoder instead of standard Transformer (ablation) | scBIG (GRC fixed) · Norman additive split | rho-delta: 0.8348 -> 0.8496 | Improved |
| use balanced OT clustering instead of unbalanced K-means (ablation) | scBIG · Norman additive split | rho-delta D: 0.7765 -> 0.8230 | Improved |
| keep stable GRC modules instead of updating clusters by perturbation (ablation) | scBIG · Norman additive split | rho-delta D: 0.8065 -> 0.8230 | Improved |
| add PPI structure prior (ablation) | scBIG · Norman additive split | rho-delta D: 0.7988 -> 0.8230 | Improved |
| add GeneCompass embeddings to clustering (ablation) | scBIG · Norman additive split | rho-delta: 0.8326 -> 0.8496 | Improved |
| add both L_corr and L_pathway losses (ablation) | scBIG · Norman additive split | rho-delta: 0.8242 -> 0.8496 | Improved |
| add L_corr alone (ablation) | scBIG · Norman additive split | rho-delta: 0.8242 -> 0.8306 | Improved |
Closest papers by meaning
- PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis (NeurIPS 2025 · similarity 0.88)
- PRESCRIBE: Predicting Single-Cell Responses with Bayesian Estimation (NeurIPS 2025 · similarity 0.88)
- STRAND: Sequence-Conditioned Transport for Single-Cell Perturbations (NeurIPS 2026 · similarity 0.87)
- PerturbReason: A Knowledge-Grounded Benchmark and Framework for Cell-State–Conditioned Mechanistic Reasoning of Perturbation Effects (NeurIPS 2026 · similarity 0.86)
- PertEval-scFM: Benchmarking Single-Cell Foundation Models for Perturbation Effect Prediction (ICML 2025 · similarity 0.86)
- Contextualizing biological perturbation experiments through language (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
