Video summary

Class 11: Protein & Ligand Preparation | VNIAS Bioinformatics Internship 2026

Main summary

Key takeaways

Science and Nature

Why protein preparation matters for docking

  • Correct protein structure preparation is required to reliably predict ligand–protein interactions.
  • Emphasized concept: “garbage in, garbage out” → inaccurate structures lead to unreliable docking results.
  • Goal: produce clean receptor (protein) and clean ligand structures with correct file formats for docking in AutoDock Vina.

Protein structure sources and initial processing

Protein structure download

  • Proteins are downloaded as PDB files from the RCSB PDB.

Protein cleaning steps

(Conceptual list mentioned; practical steps performed later)

  • Remove water molecules
  • Remove the crystallized ligand (co-crystallized inhibitor)
  • Add missing hydrogens (polar hydrogens)
  • Keep changes relevant to docking, depending on the research objective
  • Choose which chains to keep:
    • all chains,
    • a subset, or
    • a single chain containing the binding pocket

Quality checks during cleaning

  • Check and handle:
    • missing residues
    • alternate conformations
    • non-standard residues/atoms

Purpose: obtain a receptor suitable for docking.

File formats required for docking (PDB → PDBQT)

  • After cleaning, proteins and ligands must be converted to PDBQT.
  • PDBQT encodes:
    • geometry
    • docking-relevant atom typing/charges
  • Both proteins and ligands use PDBQT as input for docking tools.

Grid box definition (search space for docking)

  • A grid box definition is required to specify the docking search region in AutoDock Vina.
  • The grid box must cover:
    • the binding pocket / active site you want to target, or
    • if doing blind/global docking, an enlarged region covering most/all of the protein.
  • GUI workflow referenced:
    • determine center coordinates: centerX, centerY, centerZ
    • determine box sizes: sizeX, sizeY, sizeZ
    • tune grid spacing/size as needed

Ligand preparation concepts

Input format and preprocessing

  • Ligand inputs often come in SDF and must be cleaned and converted.

Flexibility and rotatable bonds

  • Ligand PDBQT/processing uses rotatable bonds to represent conformational flexibility.
  • More rotatable bonds → more flexible ligand → slower docking / larger search.

Redocking and validation criterion (conceptual QC step)

Redocking definition

  • Take the co-crystallized ligand from the PDB complex.
  • Dock it back into the same prepared receptor binding site.
  • Compare the redocked pose to the original crystallized complex.

Validation metric

  • RMSD should be < 2 Å for best validation.
  • RMSD > 3 Å suggests potential problems in receptor/ligand preparation.

Common docking/troubleshooting issues (conceptual list)

  • If AutoDock Vina crashes or produces errors:
    • recheck grid box settings and ensure receptor/ligand preparation is correct
  • If PDBQT ligand has zero atoms:
    • indicates invalid ligand conversion (e.g., bad SMILES or faulty SDF processing)
  • If no output file:
    • possible causes include docking command/preparation issues, or disk space limitations
  • If poses are not found / docking fails:
    • increase exhaustiveness (e.g., up to 16)
    • possibly reduce grid box size
  • If torsion/rotatable-related torsion equals zero for ligands:
    • check ligand SMILES correctness / verify SMILES input

Docking configuration parameters and batch setup (workflow)

AutoDock Vina configuration file

A config file (e.g., config*.txt / configx) typically includes:

  • Receptor: receptor.pdbqt
  • Grid: center_x, center_y, center_z and size_x, size_y, size_z
  • exhaustiveness (search thoroughness)
  • energy_range (mentioned as 3 or 4)
  • num_modes (number of output poses; commonly 1, can be 10+)
  • output location (results folder)

Multi-ligand docking workflow

  • Dock many ligands (e.g., a library of ~100) against the same receptor in batches.

Practical pipeline demonstrated (high-level methodology)

Protein pipeline

  • Download PDB structure (example used: COVID-19 main protease complex with inhibitor N3)
  • Remove waters and unwanted components
  • Add hydrogens and compute charges
  • Assign docking atom types (AutoDock4 atom typing)
  • Export to receptor.pdbqt

Ligand pipeline

  • Start with ligand library in SDF
  • Clean/preprocess using RDKit-based Python scripts:
    • normalization/sanitization/standardization
    • deduplication
    • fix issues in ligand files
  • Convert to 3D SDF
  • Convert to PDBQT using Open Babel:
    • split into separate PDBQT files per ligand (e.g., ligand_1.pdbqt … ligand_99.pdbqt)

Grid and config

  • Use a docking tool GUI (AutoDock tools) to generate the grid box
  • Save config for AutoDock Vina runs

Tool installation note

  • AutoDock tools are not web-based; they must be installed (e.g., MGLTools referenced).
  • executable tools can be copied to the working directory for access.

Researchers / sources featured

  • RCSB PDB (protein structure repository)
  • UniProt (mentioned as used earlier for protein structure analysis)
  • RDKit (Python scripts for ligand cleaning)
  • Open Babel (SDF/3D SDF → PDBQT conversion and splitting)
  • AutoDock Vina (docking engine targeted)
  • AutoDock Tools / MGLTools (receptor/ligand preparation, PDBQT creation, grid/config workflow)
  • NCBI / COVID-19 context via RCSB example:
    • example structure described as “COVID-19 main protein” in complex with inhibitor N3
    • organism listed as SARS-related coronavirus construct

Original video