Fragment-Based Drug Design (FBDD)
Purpose
Design and analyze fragment libraries, compute ligand efficiency metrics,
perform fragment docking, and execute fragment-to-lead elaboration
(growing, linking, merging) with computational support.
When to Use This Skill
- Building or filtering a fragment library
- Computing LE/LLE/LLEAT efficiency metrics
- Docking fragments into a target (weak binding, requires special settings)
- Growing a fragment hit toward lead-like compounds
- Merging two fragment hits sharing a common substructure
- Analyzing X-ray fragment screening data
Reference Files
| File |
Content |
references/fbdd-theory.md |
Fragment rules (Rule of 3), LE/LLE/LLEAT/BEI/SEI, Hann complexity model, fragment-to-lead strategies (grow/link/merge), success stories |
references/fragment-library.md |
Library design: RDKit filters (Ro3/PAINS/flatness/rigidity), 3D sp3 character, commercial sources, diversity selection, quality checks |
references/fragment-docking.md |
Low-MW docking pitfalls, Vina fragment settings, ROCS shape screening, Smina fragment mode, pose clustering, hotspot validation |
references/fragment-growing.md |
Scaffold growing (R-group enumeration, MMPA vectors), fragment merging (MCS-based), FBDD-aware REINVENT, SynthesizabilityOracle, elaboration scoring |
references/efficiency-metrics.md |
LE/LLE/LLEAT/BEI/SEI formulas, efficiency evolution plots, Abad-Zapatero plots, LELP, GE (group efficiency), metric-driven SAR |
Quick Routing
"Build a fragment library" → fragment-library.md
"Dock fragments into my target" → fragment-docking.md
"I have a fragment hit, want to grow it" → fragment-growing.md
"Track efficiency as I optimize" → efficiency-metrics.md
"What makes a good fragment?" → fbdd-theory.md
Core Concept: Rule of 3
| Property |
Fragment (Ro3) |
Lead-like |
Drug-like (Ro5) |
| MW |
≤ 300 Da |
≤ 400 Da |
≤ 500 Da |
| cLogP |
≤ 3 |
≤ 4 |
≤ 5 |
| HBD |
≤ 3 |
≤ 4 |
≤ 5 |
| HBA |
≤ 3 |
≤ 8 |
≤ 10 |
| PSA |
— |
≤ 120 Ų |
— |
| Rotatable bonds |
≤ 3 |
≤ 7 |
≤ 10 |
Minimal LE Calculation
def ligand_efficiency(pIC50, n_heavy_atoms):
"""LE = ΔG / HAC ≈ 1.37 * pIC50 / HAC (kcal/mol per heavy atom)"""
return 1.37 * pIC50 / n_heavy_atoms
# Good fragment: LE ≥ 0.3 kcal/mol/HA
# Drug-like optimum: LE ≥ 0.3 (maintain or improve during optimization)
Integration with ALKYL Skills
- Fragment docking:
docking skill (Vina/Gnina, lower exhaustiveness ok)
- Fragment diversity:
chem_diversity.py (MaxMin)
- Fragment filtering:
chem_filter.py, chem_batch.py
- Growing enumeration:
chem_react.py (ReactionFromSmarts)
- MMPA elaboration:
mmpa skill
- Generative growing:
generative-design skill (REINVENT with fragment constraint)
- 3D visualization:
py3Dmol skill
1---2name: fbdd3description: Fragment-Based Drug Design (FBDD)4---56# Fragment-Based Drug Design (FBDD)78## Purpose9Design and analyze fragment libraries, compute ligand efficiency metrics,10perform fragment docking, and execute fragment-to-lead elaboration11(growing, linking, merging) with computational support.1213## When to Use This Skill14- Building or filtering a fragment library15- Computing LE/LLE/LLEAT efficiency metrics16- Docking fragments into a target (weak binding, requires special settings)17- Growing a fragment hit toward lead-like compounds18- Merging two fragment hits sharing a common substructure19- Analyzing X-ray fragment screening data2021## Reference Files2223| File | Content |24|------|---------|25| `references/fbdd-theory.md` | Fragment rules (Rule of 3), LE/LLE/LLEAT/BEI/SEI, Hann complexity model, fragment-to-lead strategies (grow/link/merge), success stories |26| `references/fragment-library.md` | Library design: RDKit filters (Ro3/PAINS/flatness/rigidity), 3D sp3 character, commercial sources, diversity selection, quality checks |27| `references/fragment-docking.md` | Low-MW docking pitfalls, Vina fragment settings, ROCS shape screening, Smina fragment mode, pose clustering, hotspot validation |28| `references/fragment-growing.md` | Scaffold growing (R-group enumeration, MMPA vectors), fragment merging (MCS-based), FBDD-aware REINVENT, SynthesizabilityOracle, elaboration scoring |29| `references/efficiency-metrics.md` | LE/LLE/LLEAT/BEI/SEI formulas, efficiency evolution plots, Abad-Zapatero plots, LELP, GE (group efficiency), metric-driven SAR |3031## Quick Routing3233**"Build a fragment library"** → `fragment-library.md`3435**"Dock fragments into my target"** → `fragment-docking.md`3637**"I have a fragment hit, want to grow it"** → `fragment-growing.md`3839**"Track efficiency as I optimize"** → `efficiency-metrics.md`4041**"What makes a good fragment?"** → `fbdd-theory.md`4243## Core Concept: Rule of 34445| Property | Fragment (Ro3) | Lead-like | Drug-like (Ro5) |46|----------|---------------|-----------|-----------------|47| MW | ≤ 300 Da | ≤ 400 Da | ≤ 500 Da |48| cLogP | ≤ 3 | ≤ 4 | ≤ 5 |49| HBD | ≤ 3 | ≤ 4 | ≤ 5 |50| HBA | ≤ 3 | ≤ 8 | ≤ 10 |51| PSA | — | ≤ 120 Ų | — |52| Rotatable bonds | ≤ 3 | ≤ 7 | ≤ 10 |5354## Minimal LE Calculation5556```python57def ligand_efficiency(pIC50, n_heavy_atoms):58 """LE = ΔG / HAC ≈ 1.37 * pIC50 / HAC (kcal/mol per heavy atom)"""59 return 1.37 * pIC50 / n_heavy_atoms6061# Good fragment: LE ≥ 0.3 kcal/mol/HA62# Drug-like optimum: LE ≥ 0.3 (maintain or improve during optimization)63```6465## Integration with ALKYL Skills66- Fragment docking: `docking` skill (Vina/Gnina, lower exhaustiveness ok)67- Fragment diversity: `chem_diversity.py` (MaxMin)68- Fragment filtering: `chem_filter.py`, `chem_batch.py`69- Growing enumeration: `chem_react.py` (ReactionFromSmarts)70- MMPA elaboration: `mmpa` skill71- Generative growing: `generative-design` skill (REINVENT with fragment constraint)72- 3D visualization: `py3Dmol` skill
Kdevos12