DeepPurpose-MD: An End-to-End Colab-Based Ligand Discovery Pipeline Integrating Docking, Molecular Dynamics, and Deep Learning

We present DeepPurpose-MD v2.0, a unified and fully executable pipeline for ligand discovery that now seamlessly integrates deep learning–based target screening, molecular docking, and GPU-accelerated molecular dynamics (MD) simulation—all within an interactive Google Colab environment. This work describes an in-silico computational pipeline and does not provide medical advice, diagnosis, or treatment recommendations. New in v2.0: The pipeline fully supports RNA targets (including models from simRNA and others), with RNA-specific force fields, automated preprocessing, and new advanced analysis modules for allosteric and mechanistic insights—including backbone exposure (sasa proxy), Watson–Crick base pair tracking, ligand–RNA contact mapping, carboxylate–OP1/OP2 binding, and π–π stacking analysis. Users can now analyze both protein and RNA targets, and can compare ligand-bound (“holo”) and ligand-free (“apo”) simulations in a single run. All post-MD analysis is driven by an interactive, code-free UI with toggle controls—enabling rapid and reproducible trajectory exploration. Outputs include publication-ready figures, CSVs, and ready-to-visualize structures. DeepPurpose-MD features a customized fork of DeepPurpose enhanced with ProtTrans embeddings and PCA compression, achieving a Pearson correlation of 0.80 on the BindingDB COVID dataset. With modular automation, robust error handling, and complete transparency, this workflow bridges wet-lab and dry-lab efforts, making it ideal for both novice and advanced users in cheminformatics, structural biology, and AI-driven ligand design.