Method and system for predicting RNA-small molecule drug binding affinity

Abstract

This invention discloses a method and system for predicting RNA-small molecule drug binding affinity, belonging to the fields of bioinformatics and computer-aided drug design technology. The method acquires RNA sequence, SMILES representation of the small molecule drug, RNA secondary structure, three-dimensional conformation of RNA and the small molecule drug, and binding affinity annotation information. Features of RNA and the small molecule drug are extracted from three dimensions: one-dimensional sequence, two-dimensional topology, and three-dimensional conformation, with pharmacophore features of the small molecule drug integrated into the three-dimensional feature representation. Six types of features are then input into a GRU multi-view, multi-modal cross-interactive fusion network for fusion, and initial modal information is preserved using residual connections. Finally, the predicted binding affinity value is output through a regression layer. This method does not rely on the three-dimensional structure of the RNA-small molecule complex and can be used for virtual screening and optimization of RNA-targeted small molecule drugs.

Images