ml-docking-rescoring
starsforks
How to install ml-docking-rescoring?
gptomics/bioskills/ml-docking-rescoring
$npx -y skills add GPTomics/bioSkills --skill ml-docking-rescoringPerforms ML-based protein-ligand pose prediction and scoring using DiffDock-L (diffusion-based), Boltz-1 / Boltz-2 (foundation model with affinity), Chai-1, AlphaFold3 ligand, EquiBind, TANKBind, NeuralPLexer, and hybrid workflows (DiffDock pose + GNINA rescore + PoseBusters QC).
SKILL.md
| 1 | ## Version Compatibility |
| 2 | |
| 3 | Reference examples tested with: DiffDock-L (Corso et al. 2024), Boltz-1 1.0+, Boltz-2 (Passaro et al. 2025), Chai-1 0.4+, AlphaFold 3 (DeepMind), EquiBind, TANKBind, GNINA 1.1+, and PoseBusters 0.6+. |
| 4 | |
| 5 | Before using code patterns, verify installed versions match. If versions differ: |
| 6 | - Python: `pip show <package>` then `help(module.function)` to check signatures |
| 7 | - CLI: `diffdock --version`; `boltz --version` |
| 8 | |
| 9 | If code throws ImportError, AttributeError, or TypeError, introspect the installed |
| 10 | package and adapt the example to match the actual API rather than retrying. |
| 11 | |
| 12 | # ML Docking and Rescoring |
| 13 | |
| 14 | Use machine-learning models for protein-ligand pose prediction and affinity scoring. Foundation models such as AlphaFold 3, Boltz, and Chai-1 handle protein-ligand complex prediction, while DiffDock-L extends the original DiffDock method for ligand-pose sampling (Corso et al. 2023, 2024). Boltz-2 reports affinity prediction approaching physics-based free-energy methods on its evaluated benchmarks at substantially lower computational cost. Physical plausibility remains a separate requirement: on the PoseBusters Benchmark, the original DiffDock produced a correct and physically valid pose for 12% of complexes, compared with 58% for Vina and 55% for GOLD (Buttenschoen et al. 2024). Use ML sampling with independent scoring and physical validation rather than treating model confidence as sufficient. |
| 15 | |
| 16 | For classical docking, see `chemoinformatics/virtual-screening`. For pose validation (PoseBusters), see `chemoinformatics/pose-validation`. For free-energy calculations (post-docking), see `chemoinformatics/free-energy-calculations`. For PROTAC ternary complex prediction, see `chemoinformatics/protac-degraders`. |
| 17 | |
| 18 | ## ML Docking Method Taxonomy |
| 19 | |
| 20 | | Tool | Approach | Speed | Strength | Fails when | |
| 21 | |------|----------|-------|----------|------------| |
| 22 | | DiffDock-L (Corso et al. 2024) | Equivariant diffusion | GPU; hardware-dependent | Diverse pose sampling for cross-docking | Requires physical validation; OOD risk | |
| 23 | | Boltz-1 (Wohlwend et al. 2024) | AlphaFold-style foundation | GPU; hardware-dependent | Full complex prediction | Confidence is not affinity or physical validation | |
| 24 | | Boltz-2 (Passaro et al. 2025) | Boltz-1 + affinity module | GPU; hardware-dependent | Joint pose and affinity triage | Benchmark- and chemotype-dependent accuracy | |
| 25 | | Chai-1 (Chai Discovery 2024) | AlphaFold-style + language model | GPU; hardware-dependent | Open-weight complex prediction | Validate ligands and cofactors independently | |
| 26 | | AlphaFold 3 (Abramson et al. 2024) | Foundation model | Local code/weights or public server | Complex prediction with proteins and ligands | Server and local distributions have different terms and limits | |
| 27 | | EquiBind | Equivariant single-shot | <1s GPU | Fast pose | Lowest accuracy on PoseBusters | |
| 28 | | TANKBind | Distance + classifier | <1s GPU | Fast pose + score | Geometric inconsistency | |
| 29 | | NeuralPLexer | E3-equivariant generative model | GPU; hardware-dependent | Protein-ligand structure prediction | Validate geometry and confidence on the target domain | |
| 30 | | Glide (Schrödinger) | Grid-based docking and empirical scoring | License and hardware-dependent | Commercial docking workflow | License cost | |
| 31 | | GNINA 1.1 CNN | Classical sampling + CNN scoring | GPU; hardware-dependent | CNN-assisted pose ranking | Validate transfer to the target and chemotype | |
| 32 | |
| 33 | **Decision:** For pose prediction when the complex structure must also be predicted, benchmark an open model such as Boltz or Chai-1 on target-relevant controls. For a known holo receptor, DiffDock-L sampling followed by GNINA rescoring and PoseBusters checks is one auditable hybrid option. Compare it with an appropriate classical-docking baseline rather than assuming one workflow is universally superior. |
| 34 | |
| 35 | ## Candidate Workflows to Benchmark by Scenario |
| 36 | |
| 37 | | Scenario | Recommended workflow | |
| 38 | |----------|---------------------| |
| 39 | | Known holo, need fast pose | GNINA classical | |
| 40 | | Apo or AF-predicted protein, need pose | Boltz-1 or Chai-1 | |
| 41 | | Cross-docking + scaffold hopping | DiffDock-L + GNINA rescore + PoseBusters | |
| 42 | | Affinity prediction (replace FEP first-pass) | Boltz-2 affinity module | |
| 43 | | Ultralarge library (1M+) | Vina pre-filter -> GNINA on top 1% -> Boltz-2 on top 0.1% | |
People who liked this also viewed
6 picksType
- find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 2026
- grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026
- vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026
- improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026
- setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026
- triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026
