Supporting Data for "HelixDTA: Full-length sequence–structure learning for robust and interpretable drug–target affinity prediction"

This Zenodo record provides the supporting data files used for the HelixDTA drug–target affinity prediction study. The archive contains five files: davis_dataset_cleaned.csv Cleaned Davis benchmark dataset used as one of the primary input files for HelixDTA. kiba_dataset_cleaned.csv Cleaned KIBA benchmark dataset used as another primary input file for HelixDTA. protein_to_davis.7zCompressed archive containing the protein structure files corresponding to the Davis dataset. After extraction, the files should be placed under the Davis protein structure directory, for example: protein_to_kiba.7zCompressed archive containing the protein structure files corresponding to the KIBA dataset. After extraction, the files should be placed under the KIBA protein structure directory, for example: FDA_Approved_Drug_Library.csv FDA-approved drug library used for the downstream drug repurposing analysis. This file provides the candidate compound library for applying the trained HelixDTA model to identify potential drug–target interaction candidates among approved drugs. Usage To reproduce the dataset preparation used in HelixDTA (Github: HelixDTA: A deep learning framework to predict drug-target affinity): Download all files from this Zenodo record. Place davis_dataset_cleaned.csv and kiba_dataset_cleaned.csv in the dataset path specified in main.py. Extract protein_to_davis.7z and protein_to_kiba.7z. Move the extracted protein structure files into the corresponding dataset folders: Dataset/├── davis/│ └── protein/└── kiba/ └── protein/