Service

Docking

Multi-engine structure-based docking and rescoring that prioritizes physically plausible binding modes.
The core workflow combines Vina/Smina sampling with GNINA CNN rescoring
to reduce false positives and produce an audit-ready shortlist for downstream validation.

Stage coverage: KIOM Stage 2 (Docking & CNN rescoring)
Primary outputs: ranked CSV + clustered poses + plausibility flags
Downstream: Interaction (Stage 4) → Consensus (Stage 5) → MD Simulation

What we deliver

Ranked docking results (Vina/Smina) with multi-pose statistics and clustering
GNINA CNN rescoring outputs (CNNscore, CNNaffinity) for pose plausibility filtering
Representative binding poses (PDBQT/SDF/PDB) and a shortlist for validation

Optional: integrate ML rescoring (RF/GNN) after docking when you need rank stability across conditions/variants.

Best for

Down-selecting candidates from medium-to-large libraries
Comparative ranking across targets, variants, or assay conditions
PPI or challenging pockets requiring feasibility-driven filtering

If you do not yet have candidates, start from Service — Generation.

KIOM multi-stage pipeline overview (Stage 1–5)

Pipeline
Docking corresponds to KIOM Stage 2. Outputs are designed to feed Stage 4 (mechanistic validation) and Stage 5 (consensus).

ROI comparison: docking-only vs KIOM multi-stage validation

ROI
Multi-stage validation reduces false positives by penalizing geometrically implausible poses and improving experimental ROI.

Docking and rescoring run dashboard animation (example workflow)

Run
Example run dashboard: batch docking → rescoring → shortlist generation.

Example
Example complex visualization for pose review and mechanistic handoff.

RBD–amygdalin ligand interaction (2D schematic)

2D
2D interaction schematic (ligand-centric) for review and reporting.

Scope

Small-molecule docking

Multi-pose sampling + clustering (representatives per ligand)
Cross-engine comparison (Vina vs Smina) for rank robustness
GNINA CNN rescoring for plausibility filtering

Peptide / interface docking

Interface-focused evaluation and hotspot-aware criteria
Residue contact persistence indicators (for interface blocking)
Shortlist policy designed for downstream assays and MD validation

Inputs we need from you

Item	Examples
Receptor structure	PDB/mmCIF/AlphaFold model; chain selection; cofactors (if relevant)
Candidate library	SMILES/SDF (small molecules) or FASTA/structures (peptides)
Docking region	Pocket box (center/size) or residue-defined hotspot region
Run policy	Top-k poses, exhaustiveness, rescoring on/off, clustering rules

If the docking region is not defined, we can derive a first-pass pocket/hotspot definition from the structure.

What you get (audit trail)

Rank table: per ligand, Vina/Smina affinity, GNINA CNNscore/CNNaffinity, cluster ID
Plausibility flags: downrank indicators for geometry-infeasible poses
Pose package: representative poses + top-k pose set for review
Handoff map: which outputs feed interaction analysis, consensus ranking, and MD validation

The objective is not a single “best score.” The objective is a shortlist with physical plausibility and traceable evidence.

Workflow

Designed to produce outputs that are immediately usable for mechanistic validation, MD simulation, and experimental selection.

Prepare structures

Receptor preparation, ligand/peptide preparation, and pocket/hotspot definition.

Dock (multi-pose)

Vina/Smina sampling with top-k pose extraction, clustering, and rank stability checks.

CNN rescoring (GNINA)

Compute CNNscore/CNNaffinity and penalize geometrically implausible poses (false-positive reduction).

Shortlist & handoff

Deliver ranked shortlist with evidence and handoff sets for interaction analysis, consensus, and MD validation.

Handoff to Mechanistic Validation

Optional residue-level interaction fingerprints (ProLIF-style) can be generated for the top set.

Handoff to Consensus Ranking

Docking/CNN results can be combined with ML and selection logic to produce a final prioritized list.

Handoff to MD Simulation

MD is recommended for Top-N candidates rather than the full library to control cost and runtime.

Deliverables

Files

Ranked results table (CSV) with component scores (Vina/Smina + GNINA CNN)
Representative poses (PDBQT/SDF/PDB) + clustered pose set
Run configuration summary (box, exhaustiveness, top-k, clustering policy)
Optional: interaction summary (MD/PDF) for the shortlist

Decision support

Plausibility flags to reduce docking-only false positives
Cross-engine rank agreement indicators (Vina vs Smina)
MD handoff set recommendation (Top-N, representative poses)

Recommended next step

Standard flow after Docking:
Mechanistic validation (interaction evidence) → Consensus ranking → MD Simulation (final shortlist).

If the goal is purely triage, Docking + CNN rescoring can be used as a standalone screening service.

Optional add-ons

ML rescoring integration (RF/GNN) for rank stabilization across conditions
Residue-set enforcement (hotspot/motif engagement checks)
Variant panel docking (multiple receptors / variants for robustness)

Inquire (BD)
Back to Services
Go to Service — Generation
Go to Services — MD Simulation

LANGUAGE

CONTACT US