DockM8_v1.1_DEKOIS_benchmarking_data

Résumé fourni par la source

DockM8 — DEKOIS 2.0 benchmark data (corrected, v1.1.1)====================================================== This record contains benchmarking data for the DEKOIS 2.0(Demanding Evaluation Kits for Objective In silico Screening)dataset produced with DockM8 v1.1.1. It is intended forbenchmarking scoring functions and machine-learning models, and asa large archive of pre-docked poses generated by multiple dockingalgorithms. This is the CORRECTED release: RTMScore and GenScore were recomputedagainst a corrected whole-residue binding pocket, and the per-posescoring tables, the per-scoring-function selected poses, and theconsensus/performance results were regenerated to match. All otherscoring functions and the docking poses themselves are unchangedfrom the original run. Coverage: 79 targets, docked with 5 programs(gnina, plants, qvina2, qvinaw, smina). Single download: DEKOIS.tar.bz2 Archive layout (nested compression)----------------------------------- The files are NOT directly browsable — the download is a compressedarchive of compressed archives of gzipped SDFs. Use the`extract_zenodo.py` helper from the DockM8 repository to unpack them. DEKOIS.tar.bz2 # outer archive (this download)└── phase2/ └── [target].tar.bz2 # one inner archive per DEKOIS target └── [target]/ ├── [target]_activity_data.csv # experimental active/decoy labels ├── [target]_docking_library.sdf.gz # compounds docked ├── [target]_ligand.sdf.gz # reference ligand(s) ├── [target]_pocket.pdb # docking box (unchanged) ├── [target]_pocket.pdbqt # docking box (AutoDock format) ├── [target]_protein.pdb # original structure ├── [target]_protein_prepared.pdb # prepared for docking ├── [target]_corrected_10A_pocket.pdb # corrected whole-residue pocket [v1.1.1] ├── [target]_corrected_10A_pocket.pdb.json # pocket fingerprint [v1.1.1] └── results/ ├── allposes.sdf.gz # all docked poses, all programs ├── allposes_rescored.sdf.gz # poses + every SF score (RTM/Gen corrected) ├── allposes_rescored.csv # same, as a table ├── corrected_inputs/ # controlled activity universe per program [v1.1.1] │ └── [target]_[program]_{activity_data,universe_activity,excluded}.csv ├── [program]/ # per docking engine │ ├── [program]_poses.sdf.gz # raw docked poses │ ├── [program]_poses_rescored.sdf.gz # poses + scores (RTM/Gen corrected) │ └── [program]_poses_corrected_rtm_genscore.csv # corrected RTM/Gen per pose [v1.1.1] ├── scores/ │ └── [target]_[program]_[SF]_scores.csv # best-pose score per SF ├── performance/ │ └── [target]_[program]_[SF]_performance.csv # consensus enrichment per SF └── clustering/ └── [target]_[program]_[SF]_selected_poses.sdf.gz # selected pose per compound per SF [SF] runs over each scoring function and its consensus combinations.The four corrected scoring functions are RTMScore, GenScore-scoring,GenScore-docking and GenScore-balanced. How to extract-------------- Clone the DockM8 repo (main branch) and run the helper script — itreverses all three compression layers in one command: git clone https://github.com/DrugBud-Suite/DockM8.git cd DockM8 # Put the downloaded archive in a folder, then: python -m analysis.extract_zenodo /path/to/downloads /path/to/output \ --archives DEKOIS.tar.bz2 The script will: 1. Extract the outer .tar.bz2, 2. Extract each per-target inner .tar.bz2, 3. Decompress every *.sdf.gz in place. Output layout: /path/to/output/DEKOIS_2.0x/ IMPORTANT: Do NOT attempt manual extraction with `tar -xjf` alone —you will be left with nested tarballs and thousands of unreadable.sdf.gz files. Always use the helper script.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.