Interpretable Neural ODEs for Gene Regulatory Network Discovery under Perturbations

2025-01-05

AI Summary

Existing methods for causal gene regulatory network (GRN) inference under perturbations suffer from fundamental limitations: they fail to capture nonlinear dynamics, cyclic feedback loops, and time-varying cellular states. Method: We propose a biologically interpretable Neural Ordinary Differential Equation (Neural ODE)-based framework for causal GRN inference. It models cellular state evolution as a continuous dynamical system, learns latent regulatory dynamics from perturbation response trajectories, and derives directed acyclic causal graphs via Jacobian matrix analysis. Contribution/Results: This work is the first to integrate interpretable Neural ODEs with perturbation-driven modeling, relaxing strong assumptions of linearity, stationarity, and acyclicity. By combining implicit gradient optimization with sparsity regularization, our method achieves significant improvements in trajectory prediction accuracy and GRN structure recall on both synthetic and real overexpression datasets—enabling high-fidelity, interpretable, and scalable causal inference.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Modern high-throughput biological datasets with thousands of perturbations provide the opportunity for large-scale discovery of causal graphs that represent the regulatory interactions between genes. Numerous methods have been proposed to infer a directed acyclic graph (DAG) corresponding to the underlying gene regulatory network (GRN) that captures causal gene relationships. However, existing models have restrictive assumptions (e.g. linearity, acyclicity), limited scalability, and/or fail to address the dynamic nature of biological processes such as cellular differentiation. We propose PerturbODE, a novel framework that incorporates biologically informative neural ordinary differential equations (neural ODEs) to model cell state trajectories under perturbations and derive the causal GRN from the neural ODE's parameters. We demonstrate PerturbODE's efficacy in trajectory prediction and GRN inference across simulated and real over-expression datasets.

Problem

• Gene Interaction
• Dynamic Systems
• Network Construction

Innovation

• PerturbODE
• Neural Differential Equations
• Gene Interaction Prediction