Dissociation constant prediction method based on coarse-grained protein structure

The invention discloses a method for predicting a dissociation constant pKa based on a coarse-grained protein structure. Different from a traditional protein residue pKa value prediction method based on a full-atom structure model, the method can predict the pKa value based on a coarse-grained protein structure model. According to the technical scheme, the method comprises the steps that S1, whether a coarse-grained protein structure model is a single-particle model or a double-particle model is judged; s2, setting a surrounding environment range cutoff distance of a to-be-predicted central residue, and carrying out information interception on residues which are away from the central residue CA within the cutoff distance range; s3, converting the information intercepted in the step S2 into a data structure form of a residue pair, and sending the data structure form into a neural network model for training; and S4, predicting the dissociation constant by using the trained model. The objective of the invention