D-LIM: A neural network for interpretable gene–gene interactions
PLOS Computational Biology · 23.03.2026 · 22 (3) · e1014107
DOI: 10.1371/journal.pcbi.1014107 ↗Kurz gefasst
Fachartikel in der Zeitschrift PLOS Computational Biology, online erschienen am 23.03.2026. Verfasst von Shuhui Wang und 3 weiteren Autor:innen. Beteiligte Einrichtung: Laboratoire de Biophysique et Evolution, UMR CNRS-ESPCI 8231 Chimie Biologie Innovation, PSL University, Paris, France. Hauptthemen laut MEDLINE: Epistasis, Genetic, Neural Networks, Computer. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Agence Nationale de la Recherche.
Abstract (Originalsprache)
Recent advances in gene editing can produce large genotype–fitness maps for targeted genes, yet predicting the effects of mutations between genes remains challenging. Indeed, biochemical models require knowledge of underlying parameters and interactions, whereas machine learning methods typically lack interpretability, as they do not link model parameters to biological quantities. We introduce D-LIM, a neural network that infers low-dimensional fitness landscapes directly from mutation–fitness data. The distinctive feature of D-LIM is that it assumes genes act through independent gene-specific molecular phenotypes whose nonlinear interactions determine fitness. When this assumption holds, the model yields accurate predictions and interpretable effective phenotypes. Conversely, failure reveals that a low-dimensional model is insufficient. Applied to deep mutational scanning of metabolic pathways, protein–protein interactions, and yeast environmental adaptation, D-LIM achieves state-of-the-art predictive accuracy. The inferred phenotype–fitness landscapes reveal whether epistatic interactions can be captured by a low-dimensional continuous model and identify potential trade-offs. Moreover, D-LIM estimates mutational effects on the effective phenotypes, enabling weak extrapolation beyond the training domain. D-LIM demonstrates how simple structure constraints in a neural network can help inference and hypothesis generation in biology.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Laboratoire de Biophysique et Evolution, UMR CNRS-ESPCI 8231 Chimie Biologie Innovation, PSL University, Paris, France.
- LAMSADE, Universite Paris-Dauphine, PSL University, Paris, France.
- Laboratoire de Biologie Structurale de la Cellule (BIOC), CNRS, Ecole Polytechnique, Institut Polytechnique de Paris, Palaiseau, France.
ORCID-Profile: Philippe Nghe, Vaitea Opuu
Förderung
- Agence Nationale de la Recherche
Zitieren
APA
Wang, S., Allauzen, A., Nghe, P., & Opuu, V. (2026). D-LIM: A neural network for interpretable gene–gene interactions. PLOS Computational Biology, 22(3), e1014107. https://doi.org/10.1371/journal.pcbi.1014107
Vancouver
Wang S, Allauzen A, Nghe P, Opuu V. D-LIM: A neural network for interpretable gene–gene interactions. PLOS Computational Biology. 2026;22(3):e1014107. doi:10.1371/journal.pcbi.1014107
BibTeX
@article{Wang2026,
author = {Wang, Shuhui and Allauzen, Alexandre and Nghe, Philippe and Opuu, Vaitea},
title = {D-LIM: A neural network for interpretable gene–gene interactions},
journal = {PLOS Computational Biology},
year = {2026},
volume = {22},
number = {3},
pages = {e1014107},
doi = {10.1371/journal.pcbi.1014107}
}
EUVIMED