KinSPEC: A Kinase Sequence-Based Prior-Guided Explainable Framework for Predicting Compound-Protein Interactions (Data)

VSDS_truedecoy_integrated_data.csv: VSDS compound–kinase active/decoy data containing compound IDs, SMILES, UniProt IDs, KLIFS pocket sequences, activity labels, original IDs, and 4,096-bit Morgan fingerprints. VSDS_truedecoy_integrated_data_with_protein_seq.csv: VSDS compound–kinase active/decoy data with full protein sequences added. VSDS_truedecoy_morgan_one_hot.csv: Headerless 4,096-column Morgan fingerprint matrix for the VSDS data. VSDS_truedecoy_LSAE_latent_top95_features.csv: 466-dimensional LSAE molecular feature matrix for the VSDS data. KIBA_SPECTRA_VSDS_10_50_from_merged_remove_overlap_data_t0.80_cleavage_spectra_seen_target.csv: Filtered KIBA compound–kinase data for targets seen in the KinSPEC training data, including KIBA scores, SMILES, protein sequences, KLIFS sequences, and Morgan fingerprints. KIBA_10_50_from_merged_seen_target_LSAE_latent_top95_features_2.csv: 466-dimensional LSAE molecular feature matrix for the seen-target KIBA data. KIBA_SPECTRA_VSDS_10_50_from_merged_remove_overlap_data_t0.80_cleavage_spectra_seen_target_mol_deepcoy_augment.csv: Seen-target KIBA compound–kinase data augmented with DeepCoy molecular decoys. KIBA_10_50_from_merged_seen_target_mol_deepcoy_augment_LSAE_latent_top95_features_2.csv: 466-dimensional LSAE molecular feature matrix for the DeepCoy-augmented seen-target KIBA data. KIBA_SPECTRA_VSDS_10_50_from_merged_remove_overlap_data_t0.80_cleavage_spectra_unseen_target.csv: Filtered KIBA compound–kinase data for targets not seen in the KinSPEC training data. KIBA_10_50_from_merged_unseen_target_LSAE_latent_top95_features_2.csv: 466-dimensional LSAE molecular feature matrix for the unseen-target KIBA data. KIBA_SPECTRA_VSDS_10_50_from_merged_remove_overlap_data_t0.80_cleavage_spectra_unseen_target_mol_deepcoy_augment.csv: Unseen-target KIBA compound–kinase data augmented with DeepCoy molecular decoys. KIBA_10_50_from_merged_unseen_target_mol_deepcoy_augment_LSAE_latent_top95_features_2.csv: 466-dimensional LSAE molecular feature matrix for the DeepCoy-augmented unseen-target KIBA data. Karaman_origin.csv: Original wide-format Karaman kinase-profiling matrix containing compound names, SMILES, kinase names, UniProt IDs, and bioactivity values. Karaman_expand_final.csv: Long-format Karaman compound–kinase data containing bioactivity values and binary labels at 1,000 nM and 100 nM thresholds. Karaman_Kinase_selectivity_filtered_withseq.csv: Filtered Karaman kinase-selectivity data containing SMILES, kinase names, UniProt IDs, bioactivity labels, Morgan fingerprints, and protein sequences. Karaman_10_50_filtered_morgan_one_hot_from_merge_2.csv: Headerless 4,096-column Morgan fingerprint matrix for the filtered Karaman data. Karaman_10_50_filtered_SAE_latent_top95_features_from_merge_2.csv: 466-dimensional molecular feature matrix for the filtered Karaman data. Karaman_sim0.5_SAE_latent_top95_features_from_merge.csv: 466-dimensional molecular feature matrix for the Karaman-specific training pool filtered using the sim0.5 condition.