MNAF-Net

Step1:Download the public dataset from https:\/\/github.com\/luoyunan\/DTINet.Step2:Extraction of 2D Molecular Graphs for DrugsSMILES strings were converted into 2D molecular graph structures using the RDKit toolkit. The visual representations of these structures were stored in graph files, accompanied by the generation of adjacency matrices and edge adjacency matrices.The visual representations of these structures are stored in \u201cgraph\u201d files.Step3:Feature Extraction from 2D Molecular GraphsA Graph-Cross Modal Attention Network (GCMANet) was designed to iteratively update node and edge features, progressively aggregating local structural information.Step4:3D Graph Construction for Target ProteinsThe FASTA sequences of target proteins were mapped to UniProt IDs using the UniProt database.AlphaFold II\/III or Protein Data Bank Database\uff08PDB\uff09 were employed to predict 3D protein structures. Predicted structures were stored as PDB files.The generated PDB structures are stored in \u201cgraph\u201d files.Step5:Feature Extraction from 3D Protein StructuresAtomic-level features:3D coordinate matrix of C\u03b1 atoms\u3001Residue distance matrix\u3001Contact matrixResidue-level features:Amino acid types encoded via one-hot vectors\u3001Physicochemical properties.Geometric and topological descriptors:Cosine similarity between residue pairs\u3001Dihedral angles.Structural features were systematically stored in \u201cfeature\u201d files.