df.parquet.gzip is processed raw data. This has some genes but is not fully overwrapped with training/test data drug2actions.csv is processd data that has information about drugs' effectiveness. drug2gene_from_db.csv is a drug-target(gene) interaction from 5 databases. drug2target_from_drugbank.csv is a drug-target(gene) interaction from drugbank. gene2go.csv is a gene-gene ontology relation data. Gene ontology is a hierarchal network that shows gene functions such as "Biological functions (GO:000528)". This has gene name, ontology, role, and explanation. gene2go_train.csv is filtered data that is included in training data. gene2go_onehot.csv is a dummy table, indexes are genes and columns are gene ontology. If they have a relations return True. merged_drug_gene_actions.csv is a whole dataset that contains drug2action, and drug2target from drug bank. drug2gene.csv is a drug-target(gene) combination. Known drug-gene interaction is defined. The combination is already standardized. smiles2fingerprint.csv SMILES (chemical vector structure) is converted to one hot vector derived from chemical taxonomy. For instance, if the chemical structure has "CH3" the vector has one for "CH3" columns. This is generated by RDKit