Fragment-Based Drug Design
Discover molecular fragments, grow them into drug-like compounds, and validate binding through an iterative FBDD pipeline.
Deprecated
This workflow is deprecated as of 2026-10-18: Fragment Library and Fragment Growing, the tools it is built on, are no longer available on the platform, and neither is FRoGS. This page is kept for reference only.
Fragment-Based Drug Design
Fragment-based drug design (FBDD) starts with small, low-complexity molecular fragments and systematically elaborates them into potent, drug-like compounds. This approach explores chemical space more efficiently than high-throughput screening because fragments sample a larger proportion of possible binding interactions per atom.
Overview
The FBDD strategy relies on identifying fragments (typically 150-300 Da) that bind weakly but with high ligand efficiency to a target protein. These fragments are then grown, merged, or linked to generate larger molecules with improved potency while maintaining favorable physicochemical properties. This workflow uses the Fragment Library for initial screening, Fragment Growing for systematic elaboration, and FRoGS for pathway-level validation of generated compounds to ensure biological relevance.
Tool Chain
- Fragment Library (retired) -- High-performance fragment search across curated databases
- Fragment Growing (retired) -- AI-driven fragment elaboration with quality filters
- FRoGS -- Gene signature analysis for mechanism validation
Prerequisites
- A validated protein target with known or predicted binding site
- Protein structure in PDB format (experimental or ESMFold-predicted)
- Understanding of Rule of Three criteria for fragment selection
Step-by-Step Guide
Define the fragment search space
Open the Fragment Library tool and define your search parameters. For classical FBDD, apply Rule of Three filters: molecular weight below 300 Da, LogP below 3, hydrogen bond donors no more than 3, hydrogen bond acceptors no more than 3, and rotatable bonds no more than 3. Enable PAINS filtering to exclude promiscuous binders. Choose databases to search (ZINC fragments, ChEMBL fragments, or PubChem).
Screen fragments by scaffold
If you have a known pharmacophore or scaffold preference, use substructure or SMARTS pattern searches to focus the fragment set. For example, search for all fragments containing an indole core, a pyridine ring, or a specific heterocyclic motif relevant to your target class. This generates a focused library of 50-500 fragments for virtual evaluation.
Dock fragments against the target
Submit the fragment set to DiffDock for binding pose prediction against your protein structure. Fragments bind weakly, so use a lower confidence threshold than for drug-sized molecules. Evaluate binding modes by inspecting hydrogen bond networks, hydrophobic contacts, and the position of growth vectors that point toward unoccupied regions of the binding site.
Select fragments for growing
Rank fragments by a composite score of docking confidence, ligand efficiency (binding energy divided by heavy atom count), and synthetic tractability. Select 3-10 fragments with diverse scaffolds and clear growth vectors for elaboration. Prioritize fragments that make key interactions with catalytic or recognition residues in the binding site.
Grow fragments into leads
Submit selected fragments to Fragment Growing. Specify growth vectors (attachment points on the fragment where new groups will be added), choose the fragment database for sourcing elaboration groups, and apply quality filters (Lipinski compliance, PAINS exclusion, synthetic accessibility score below 4). Configure diversity controls to ensure the generated compounds explore different regions of chemical space.
Validate with gene signature analysis
For the most promising grown compounds, run FRoGS analysis to verify that the predicted biological mechanism aligns with your therapeutic hypothesis. Gene signature analysis connects compound structures to pathway-level effects, helping to confirm on-target activity and flag potential off-target liabilities before committing to synthesis.
Prioritize and export
Rank the final compound set by a multi-parameter optimization score combining predicted binding affinity, ADMET properties (run through ADMET-AI), synthetic accessibility, and mechanism confidence from FRoGS. Export the prioritized list with all supporting data for medicinal chemistry review.
Expected Outputs
- Curated fragment hit list with Rule of Three compliance metrics
- Fragment binding poses and ligand efficiency scores
- Grown compound libraries with property profiles and diversity analysis
- Gene signature validation results from FRoGS
- Multi-parameter ranked candidate list for synthesis
Tips and Best Practices
- Enforce Rule of Three filters strictly during fragment selection. Fragments that already violate drug-likeness rules leave little room for property optimization during growth.
- Evaluate ligand efficiency (LE) rather than raw binding scores when comparing fragments. A fragment with LE above 0.3 kcal/mol/heavy atom is a strong starting point.
- Specify multiple growth vectors per fragment to explore different elaboration directions. The optimal growth direction is not always obvious from a single docking pose.
- Use SAFE (Sequential Attachment-based Fragment Embedding) notation when available for precise fragment-to-molecule tracking through the pipeline.
- Plan fragment growing with synthetic accessibility in mind. Generated compounds with SA scores above 5 are likely impractical to synthesize.
- Iterate the grow-dock-filter cycle at least twice. The first round identifies promising directions; the second round refines within the most productive chemical series.
Was this page helpful?