Bioactivity deep learning for structure-free compound-protein interaction

CPI2M data for "Bioactivity deep learning for structure-free compound-protein interaction".

CPI2M_main_Ki.csv: Bioactivity data with pKi activity type. Used for model training and internal validation.

CPI2M_main_Kd.csv: Bioactivity data with pKd activity type. Used for model training and internal validation.

CPI2M_main_EC50.csv: Bioactivity data with pEC50 activity type. Used for model training and internal validation.

CPI2M_main_IC50.csv: Bioactivity data with pIC50 activity type. Used for model training and internal validation.

CPI2M_few_Ki.csv: Bioactivity data with pKi activity type. Used for external validation.

CPI2M_few_Kd.csv: Bioactivity data with pKd activity type. Used for external validation.

CPI2M_few_EC50.csv: Bioactivity data with pEC50 activity type. Used for external validation.

CPI2M_few_IC50.csv: Bioactivity data with pIC50 activity type. Used for external validation.

potency.csv: BIoactivity data with pPotency activity type. Not used currently but can be potentially adopted as classification data for customized use.

percentage.csv: BIoactivity data with Percentage Inhibition activity type. Not used currently but can be potentially adopted as classification data for customized use.

Protein_pretrained_feat.zip: pre-calculated protein feature files with UniProt ID naming. Should be unzipped before start model training with CPI2M data.

For each .csv data, columns include "smiles" (ligand SMILES), "exp_mean" (nM bioactivity), "y" (neg.log nM, final label), "cliff_mol" (whether activity cliff or not), "split" (splitting label by activity cliff), "Uniprot_id" (UniProt ID for protein), "Sequence" (wildtype sequence for protein), and "type_id" (bioactivity type token, pKi =0, pKd=1, pEC50=2, pIC50=3).

Please find the project code at https://github.com/gu-yaowen/GGAP-CPI