MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074073 A) filed by National Institute Of Technology Calicut, An Indian Institute on June 15, 2026, for A Graph Based Contrastive Learning Framework For Reducing Protein Search Space In Reverse Docking Workflow.
Inventors include Lashma Kadassery; and Gopakumar Gopalakrishnan Nair.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT A GRAPH BASED CONTRASTIVE LEARNING FRAMEWORK FOR REDUCING PROTEIN SEARCH SPACE IN REVERSE DOCKING WORKFLOW The present invention relates to systems and methods for accelerating reverse docking by reducing the protein search space using graph-based representation learning. Protein binding pockets are transformed into graph representations in which nodes correspond to atomic entities and edges represent spatial or physicochemical relationships. A graph neural network is employed to generate low-dimensional embedding vectors that encode structural and interaction-relevant characteristics of the binding pockets. The model is trained using a contrastive learning framework to learn invariant and discriminative representations of protein pockets. During inference, the embedding of a query-ligand binding environment is compared with a precomputed library of protein pocket embeddings to rank and prioritize candidate proteins. By pre-filtering proteins based on embedding similarity prior to atomistic docking simulations, the system significantly reduces computational complexity, improves scalability, and enables efficient drug repurposing across large protein datasets. FIG. 2
Disclaimer: Curated by HT Syndication.
