D-LIM Infers Low-Dimensional Gene–Gene Fitness Landscapes
Researchers (Wang et al.) publish on March 23, 2026 a neural network, D-LIM, that infers low-dimensional fitness landscapes directly from mutation–fitness data by modeling gene-specific molecular phenotypes whose nonlinear interactions determine fitness. Applied to deep mutational scanning of metabolic pathways, protein–protein interactions, and yeast environmental adaptation, D-LIM achieves state-of-the-art predictive accuracy and yields interpretable effective phenotypes; code and datasets are available on GitHub.
Key Points
- 1Introduces D-LIM neural network that infers low-dimensional phenotype–fitness landscapes from mutation–fitness data.
- 2Demonstrates state-of-the-art accuracy across metabolic, protein–protein, and yeast adaptation deep-mutational datasets.
- 3Enables interpretable effective phenotypes, detects low-dimensional epistasis, and supports weak extrapolation beyond training data.
Scoring Rationale
Novel, peer-reviewed method with usable code and broad applicability; primarily focused on genotype-to-fitness domain, limiting general ML transfer.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
