mlsanityData to 13 Oct 2026

No Foundations without Foundations -- Why semi-mechanistic models are essential for regulatory biology

arXiv 2501.19178 ↗First posted 31 Jan 2025cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

Despite substantial efforts, deep learning has not yet delivered a transformative impact on elucidating regulatory biology, particularly in the realm of predicting gene expression profiles. Here, we argue that genuine "foundation models" of regulatory biology will remain out of reach unless guided by frameworks that integrate mechanistic insight with principled experimental design. We present one such ground-up, semi-mechanistic framework that unifies perturbation-based experimental designs across both in vitro and in vivo CRISPR screens, accounting for differentiating and non-differentiating cellular systems. By revealing previously unrecognised assumptions in published machine learning methods, our approach clarifies links with popular techniques such as variational autoencoders and structural causal models. In practice, this framework suggests a modified loss function that we demonstrate can improve predictive performance, and further suggests an error analysis that informs batching strategies. Ultimately, since cellular regulation emerges from innumerable interactions amongst largely uncharted molecular components, we contend that systems-level understanding cannot be achieved through structural biology alone. Instead, we argue that real progress will require a first-principles perspective on how experiments capture biological phenomena, how data are generated, and how these processes can be reflected in more faithful modelling architectures.

Key facts. Cited by 3 papers in the map (data through 13 Oct 2026). Builds on 17 of its 75 references that are in the map. First posted on arXiv 31 Jan 2025.

Builds on

17 of 75 references.

References · 17

  1. 2018Neural Ordinary Differential Equationscited by 3,750
  2. 2022On Neural Differential Equationscited by 265
  3. 2017Permutation-based Causal Inference Algorithms with Interventionscited by 78
  4. 2018Characterizing and Learning Equivalence Classes of Causal DAGs under Interventionscited by 76
  5. 2024PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysiscited by 35
  6. 2022Learning Causal Representations of Single Cells via Sparse Mechanism Shift Modelingcited by 34
  7. 2022Active Learning for Optimal Intervention Design in Causal Modelscited by 30
  8. 2022Deep Learning of Causal Structures in High Dimensionscited by 20
  9. 2023Causal Modeling with Stationary Diffusionscited by 19
  10. 2021Near-Optimal Multi-Perturbation Experimental Design for Causal Structure Learningcited by 17
  11. 2024Benchmarking Transcriptomics Foundation Models for Perturbation Analysis : one PCA still rules them allcited by 17
  12. 2024Season combinatorial intervention predictions with Salt & Pepercited by 11
5 more
  1. 2024Learning Identifiable Factorized Causal Representations of Cellular Responsescited by 4
  2. 2023DiscoGen: Learning to Discover Gene Regulatory Networkscited by 3
  3. 2024Simulation-based Benchmarking for Causal Structure Learning in Gene Perturbation Experimentscited by 3
  4. 2022PyRelationAL: a python library for active learning research and developmentcited by 2
  5. 2016Joint Causal Inference from Multiple Contextscited by 1

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2025: 1; 2026: 2.

3 papers.

  1. 2026Chem-PerturBridge: a harmonized compendium of small molecule perturbation transcriptomic effectscited by 1
  2. 2025When three experiments are better than two: Avoiding intractable correlated aleatoric uncertainty by leveraging a novel bias--variance tradeoffcited by 0
  3. 2026An Investigation of the NeurIPS and ICML 2025 Position Trackscited by 0