Gene regulatory networks: from correlative models to causal explanations - PubMed
2 days ago
- #Representation Learning
- #Single-cell Technologies
- #Gene Regulatory Networks
- Gene regulatory networks (GRNs) connect molecular mechanisms to functional outputs in cellular behavior and tissue morphogenesis.
- Single-cell technologies provide detailed GRN descriptions but reveal overly complex systems, reducing GRNs to statistical correlations ('hairballs') lacking causal explanations.
- The article proposes using 'representation learning' to model GRNs without capturing every molecular detail.
- Three principles are advocated: mechanistic models grounded in cellular and evolutionary biology, molecular constraints to reduce solution space, and advanced experimental perturbations for model training and testing.
- The goal is to bridge the gap from abundant data to new conceptual understanding in GRN research.