fbdd - SKILL.md Agent Skill

name: fbdd description: Use for fragment-based drug design (FBDD): Rule of 3 filtering, ligand efficiency metrics (LE/LLE/BEI/LELP), fragment library design, fragment docking (Vina/Gnina), fragment growing/linking/merging strategies, and Abad-Zapatero efficiency plots.

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 Å²	—
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