Drug-target affinity predictions data for evaluation of interaction concordance index

This record contains data for repeating the results presented in "Interaction Concordance Index: Performance Evaluation for Interaction Prediction Methods". It contains ground-truth drug–target affinity values, the corresponding drug/target index pairs, the cross‑validation fold assignments used in the experiments, and prediction outputs produced by one or more learning methods.

File naming conventions- ground_truth_DATA.csv  - Real-valued affinity measurements for dataset DATA.  - One value per line; no header.- drug_target_index_DATA.csv  - Index pairs identifying the drug and target for each affinity observation in dataset DATA.  - Two nonnegative integers per line separated by comma: drug_index,target_index; no header.  - Indices are 0-based (start at 0) and refer to the DATA-specific drug and target sets.- folds_DATA.csv  - Cross-validation partition used in the experiments for dataset DATA.  - One integer in range [0,8] per line indicating the fold assignment; no header.- predictions_METHOD_CONDITION_DATA.csv  - Predicted affinity values for dataset DATA generated by learning method METHOD trained under CONDITION.  - One real-valued prediction per line; no header.

Alignment and integrity- For a given DATA, all files associated to it have exactly the same number of lines and the same ordering as ground_truth_DATA.csv, so that line i refers to the same observation across these files: the ground truth affinity value at line i corresponds to the predicted affinity value at line i, the drug/target indices at line i and is assigned to the fold at line i.- No missing lines or header rows are present in any file.

Formats and conventions- CSV with comma delimiter, UTF-8 encoding, Unix line endings.- Decimal separator: dot (e.g., 0.1234).- Whitespace is not significant; trailing/leading spaces should be ignored.- Valid value types:  - Affinity (ground truth, predictions): floating-point numbers (e.g., 1.0, 0.567, NaN not allowed).  - Indices: integers >= 0 (0-based).  - Folds: integers >= 0 (0-based).

Placeholders- DATA: identifier of a dataset (davis, metz, merget, kiba, IC, E, GPCR).- METHOD: identifier of the learning algorithm (KRLSKRG, KRLSLRG, KRLSKRL, KRLSLRL, kNN, ltr, RF, XGBoost, DDTA, FF, GT).- CONDITION: learning problem (IDIT, IDOT, ODIT, ODOT).