Method for predicting unknown small molecule ligand target spot and application of method

The invention discloses a method for predicting an unknown small molecule ligand target and an application of the method, wherein the method comprises the steps: firstly, classifying ligand-protein fragment compounds which have interaction and have the same ligand and protein fragment type into one class, and superimposing the ligand-protein fragment compounds based on a protein fragment structure to construct superimposing models of the class; clustering ligand atoms of various superposition models by adopting a Bayesian Gaussian mixture model to represent spatial distribution of the ligand atoms in the various superposition models, then inputting combined conformation data of protein fragments and unknown small molecule ligands, and classifying the combined conformation data; and counting the number of interaction pairs according to the distance between each atom of the protein fragment and each atom of the unknown small molecule ligand, then calculating the distance between the unknown small molecule ligan