跳转至

中文文档

Docking Tasks

Role

Docking tasks combine receptor, ligand, and pocket assets to predict binding poses, screen candidates, and seed FEP. The default docking engine is Uni-Dock GPU with Vina/Vinardo scoring.

Inputs

  • prepared_protein asset (protein, complex, and MD-derived md_structure assets are also accepted as receptor)
  • ligand or prepared_ligand asset
  • pocket asset, or a pocket_ensemble (dynamic pocket ensemble) asset from GROMACS pocket_discovery: with an ensemble selected, no separate receptor is required and each pocket uses its own cluster representative structure as receptor
  • Uni-Dock is a traditional docking engine and does not require neural network model files

Outputs

  • JobOut
  • result asset: includes the Uni-Dock report, pose table, worker log, and task summary.
  • 1 prepared_ligand_library / docking_pose_library: all successful poses from this task are merged into one SDF. Each SDF record keeps molecule name, SMILES, docking score, and pose index, and can be selected, sorted, exported, or reused by interaction analysis/FEP.
  • When a pocket_ensemble is used, the task fans out over "pocket x ligand" runs while still merging all poses into one SDF; each record additionally carries WA_DD_POCKET_INDEX / POCKET_LABEL / POCKET_CENTER / POCKET_CLUSTER_POPULATION properties, and the report and pose table add cross-pocket consensus ranking (best score, best pocket, and pocket-hit coverage per molecule). Interaction outputs are generated per pocket.

Workflow

  1. Import or prepare the receptor in Protein Processing, then save a pocket asset from a co-crystal ligand or a manual box.
  2. Import SDF/SMILES or draw molecules in Ligand Processing. Select the export profile by target use:
  3. Uni-Dock/Vina: add hydrogens, generate 3D conformers, assign Gasteiger charges, and prepare PDBQT.
  4. FEP/MD: keep 3D conformers and force-field handoff metadata.
  5. In Docking Tasks, select receptor, pocket, and ligand from dropdowns. The pocket dropdown accepts both standard pocket and dynamic pocket ensemble (pocket_ensemble) assets; when an ensemble is selected the receptor can be left empty. The page calls /api/v1/docking/compatibility to check whether the current combination is runnable.
  6. Choose the docking method (default: Uni-Dock GPU) and related parameters.
  7. Click Submit Docking Task. The system creates a docking job and dispatches it to the Uni-Dock worker.
  8. Track step-by-step progress in the Task Center. After completion, go to the output asset to download structures/reports, or reuse the result asset in analysis/FEP.

API / Automation

GET /api/v1/docking/compatibility
POST /api/v1/jobs
GET /api/v1/jobs?project_id=<project_id>
GET /api/v1/jobs/{job_id}/events
GET /api/v1/models/pocketxmol

For automation, call the compatibility endpoint first. Submit a docking task only when compatible=true. After completion, read output_asset_ids and download structures/reports from the asset file endpoints.

Server6 Example: 1TA2 Congeneric Library and de novo Generated Ligand Docking

This example was completed in the Example project on server6, using the 176 ligand pocket of 1TA2 thrombin.

Docking task: select receptor, pocket, ligand, and Uni-Dock parameters

Docking task: view report or enter interaction analysis after completion

Key steps:

  1. For the protein, select Example 1TA2 receptor prepared ligand-removed.
  2. For the pocket, select Example 1TA2 176 binding pocket.
  3. The ligand library can be one of:
  4. Example 1TA2 congeneric 72 analog library prepared for docking
  5. Example 1TA2 PocketXMol de novo generated ligands
  6. For the docking method, choose Uni-Dock; this example uses pose_per_ligand=3, keep_top_poses=1, cpu_threads=4, and GPU device cuda:0.
  7. After the task completes, use:
  8. View Report · N poses: open the pose table and rank by docking score.
  9. Analyze Pose Library · N poses: automatically load this task's merged SDF into the interaction analysis page.

Validation output for this run:

  • Congeneric library: 66 input ligands, 194 poses, 0 failed/skipped; the pose library asset is Example 1TA2 congeneric Uni-Dock screening result docked pose library.
  • de novo generated ligands: 23 input ligands, 69 poses, 0 failed/skipped; the pose library asset is Example 1TA2 PocketXMol de novo Uni-Dock result docked pose library.

Result interpretation:

  • A more negative docking score generally indicates better Vina scoring, but it is not experimental binding free energy.
  • A single task outputs only one merged SDF; it no longer splits each pose into a pile of assets. To compare results, sort by score in the pose table or interaction analysis and view them with single/multi-selection.
  • If View Report does not respond or reports an error, first check whether the task has produced unidock_pose_table.csv and unidock_docked_poses.sdf; these two files are the source of the report and 3D analysis.