Integrated Multi-Omics Graph Neural Networks and Molecular Simulation Pipeline for Drug Response and Mechanism Discovery
Abstract
Understanding drug response at the molecular level requires integrating heterogeneous biological data with mechanistic modeling approaches. In this project, we develop an advanced computational framework that combines multi-omics data integration using Graph Neural Networks (GNNs) with molecular docking and molecular dynamics (MD) simulations to identify drug targets, pathways, and mechanisms of action.
The core of the project is a modular multi-omics pipeline integrating transcriptomics, proteomics, CRISPR screening, and imaging data into a unified graph representation. The system employs adaptive learning strategies, uncertainty estimation, and hypothesis-driven optimization to discover biologically relevant pathways and predict drug response.
To complement data-driven modeling, structure-based computational methods, including virtual screening and MD simulations, are incorporated to enable mechanistic validation of predicted targets and interactions.
The expected outcome is a scalable and reproducible computational framework capable of predicting drug response, identifying regulatory pathways, and validating molecular mechanisms. This work contributes to computational drug discovery and precision medicine by bridging machine learning and molecular biophysics.
