MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641095329 A) filed by Vardhaman College Of Engineering on August 06, 2026, for Graph Neural Networks For Predicting Protein-Protein Interactions In Drug Discovery.
Inventors include Ms. Shaik Shahanaz; Ms. Dumpala Dhana Lakshmi; Mr. B J V Varma; Mr. Surendhar S; Ms. C Kavitha; and Dr. Viyyapu Lokeshwari Vinya.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Graph Neural Networks for Predicting Protein-Protein Interactions in Drug Discovery is the proposed invention. The invention disclosed is a Graph Neural Network based system to predict protein-protein interactions for accelerating drug discovery and therapeutic target identifications. The system combines protein sequence information, structural information, gene ontology annotations, biological pathways, diseases annotations and experimentally validated interaction data into a unified protein interaction graph. Each protein is represented by a graph node with multi-dimensional biological features, and the known interactions are weighted graph edges. We learn discriminative protein embeddings by iteratively aggregating neighbourhoods and convolving graphs to capture complex biological relationships using a Graph Neural Network. The interaction prediction module, using these embeddings, was able to recover both known and novel protein interactions with high accuracy. The predicted interaction networks are further analysed to prioritise disease associated therapeutic targets as well as for drug repurposing. An Explainable Artificial Intelligence module delivers transparent interpretation by pinpointing influential neighbouring proteins and biological pathways that contribute to predictions. The framework supports incremental graph updates enabling scalable continuous learning, and efficient integration into modern computational drug discovery pipelines, reducing experimental cost and development time.
Disclaimer: Curated by HT Syndication.
