A deep learning method for predicting interactions for intrinsically disordered regions of proteins

Note to user The files have been organized according to the directory structure that is needed for running the scripts. Disobind training set includes the train, development, and in-distribution test set. This dataset contains the following: disobind/ Zipped GitHub file. prose and ProstT5 GitHub repo used. database/ Combined_PDBs.tar.gz: a zipped file containing all PDB structures. Mapped_PDBs.tar.gz: a zipped file containing SIFTS mapping for all PDB entries. PDB_api.tar.gz: a zipped file containing entry and entity-level information from PDB API as JSON files. v_19: AF2_preds: AlphaFold-multimer predictions for OOD test set. AF3_preds: AlphaFold3 predictions for OOD test set. Target_bcmap_test_v_19.h5: target contact maps for OOD test set. Target_bcmap_train_v_19.h5: target contact maps for Disobind training set. Uniprot_seq.json: JSON file containing UniProt sequences for all UniProt accessions included in the dataset. prot_1-2_test_v_19.csv: entry IDs for all merged binary complexes in the OOD test set. prot_1-2_train_v_19.csv: entry IDs for all merged binary complexes in the Disobind training set. T5/ global-None: contains ProtT5-global embeddings for the train, development, and in-distribution test set. local-None: contains ProtT5-local embeddings for the train, development, and in-distribution test set. train_fasta_global-None_v_19.fasta: FASTA file used as input for creating ProtT5-global embeddings for train, development, and in-distribution test set. train_fasta_local-None_v_19.fasta: FASTA file used as input for creating ProtT5-local embeddings train, development, and in-distribution test set.