Protease data prediction method for catalytic reaction and application

Abstract

This invention discloses a method and application for protease data prediction for catalytic reactions. The method includes: acquiring protein sequence data, pocket masks, and molecular structure data of chemical reactions; inputting the protein sequence data into a first encoder to obtain protein feature data, and extracting pocket feature data based on the pocket mask; inputting the molecular structure data into a second encoder to obtain reaction feature data; mapping each feature data to a hyperbolic space using learnable hyperbolic mapping parameters; calculating the contrast loss value based on hyperbolic distance and the implication loss value based on geometric relationships; updating the encoder parameters and hyperbolic mapping parameters to obtain a protein prediction model; and using this model to output hyperbolic representation data of proteins and / or hyperbolic representation data of reactions for enzyme catalytic function prediction, candidate enzyme and molecular structure data screening, or enzyme sequence generation. This invention achieves accurate enzyme function prediction and zero-sample enzyme design based solely on the sequence.

Images