DYNAMITE: DYnamic biomolecular interaction Network And Machine learning Integration for Targeted Energy calculations

AI Technology: Machine Learning; Deep Learning

The DYNAMITE (DYnamic biomolecular interaction Network And Machine learning Integration for Targeted Energy calculations) project aims to revolutionize binding free energy (BFE) predictions by integrating large-scale molecular dynamics (MD) simulations with advanced machine learning (ML) methodologies. 

Building upon the BFEx (Binding Free Energy Exscalate) tool, which efficiently extracts molecular and atomic interaction data from MD trajectories, DYNAMITE will process over 16,000 MD simulations to generate a dataset comprising millions of interaction patterns. 

This extensive dataset will be curated, refined, and leveraged to train ML models, including Random Forest, XGBoost, and Transformer-based deep learning architectures, for accurate and scalable BFE predictions. 

By utilizing the computational power of EuroHPC infrastructure, the project will significantly enhance the efficiency of BFE calculations, surpassing traditional methods such as docking, free energy perturbation (FEP), and MM/GBSA. The predictive models developed will undergo rigorous benchmarking to ensure their reliability and generalizability across diverse protein-ligand systems.

Additionally, the project will develop automated, scalable workflows that streamline BFE calculations, reducing computational costs and accelerating drug discovery pipelines. 

The expected outcomes include basic & advance ML models, an extensive BFE dataset, a user-friendly BFEx tool, and new benchmarking standards for BFE prediction methods. By bridging the gap between MD simulations and ML-driven predictive analytics, DYNAMITE will set a new paradigm in computational drug discovery, enhancing precision medicine and therapeutic development strategies.