MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202621075942 A) filed by Symbiosis International Deemed University on June 18, 2026, for Graph Convolutional Network System For Backbone-Based Prediction Of Amino Acid Probabilities In Protein Sequence Design.
Inventor includes Dr. Shreyas Rajendra Hole.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: ABSTRACT GRAPH CONVOLUTIONAL NETWORK SYSTEM FOR BACKBONE-BASED PREDICTION OF AMINO ACID PROBABILITIES IN PROTEIN SEQUENCE DESIGN The present invention discloses a graph convolutional network system (100) referred to as ProtGCN for predicting amino acid probabilities from protein backbone three-dimensional coordinates. The system (100) comprises a data preprocessing module (110) for extracting backbone atom coordinates and computing pseudo-NH and pseudo-CB atoms, a graph construction module (120) for building directed N-nearest neighbor graphs using pseudo-CB distances with 4D torus dihedral angle node features and 6x6 inter-residue distance matrix edge features, a hybrid encoder network (130) combining graph convolutional layers for spatial message passing with 1D convolutional layers for 9-residue fragment local context extraction, and a prediction network (140) using 1D convolution with softmax activation outputting probability distributions over 20 amino acids. The system (100) achieves 53.56% accuracy on T500 and 47.61% on TS50, with top-3 accuracy 72.37% and top-5 accuracy 81.58%, at 0.23 seconds per protein enabling large-scale protein engineering. [
Disclaimer: Curated by HT Syndication.
