Fragment-Based Screening

A physicochemical guideline defining fragment-like chemical space. Fragments typically adhere to: molecular weight < 300 Da, clogP ≤ 3, hydrogen bond donors ≤ 3, and hydrogen bond acceptors ≤ 3. This constraint ensures high ligand efficiency and leaves ample room for subsequent chemical optimization without violating Lipinski's Rule of Five for the final lead compound.

The central metric for ranking fragment hits, normalizing binding energy by molecular size. Ligand Efficiency (LE) is calculated as the free energy of binding divided by the number of heavy atoms (non-hydrogen atoms). A typical threshold for a promising fragment is LE ≥ 0.3 kcal/mol per heavy atom. Other indices like Lipophilic Ligand Efficiency (LLE) balance potency against lipophilicity to avoid greasy, non-specific binders.

Fragments bind with weak affinity (typically KD in the μM to mM range), making them undetectable by standard biochemical assays. Detection relies on sensitive biophysical techniques:

• NMR Spectroscopy: Ligand-observed methods like WaterLOGSY and STD-NMR detect binding through changes in nuclear relaxation or magnetization transfer.
• Surface Plasmon Resonance (SPR): Measures real-time binding kinetics and affinity by detecting mass changes on a sensor chip.
• X-ray Crystallography: Provides high-resolution structural data on fragment binding poses, enabling direct structure-guided optimization.

Once validated, weak-binding fragments are evolved into potent leads through three primary strategies:

• Fragment Growing: Iteratively adding functional groups to the fragment core to probe adjacent binding pockets and increase affinity.
• Fragment Linking: Connecting two fragments that bind to proximal, non-overlapping sites on the target with a chemical linker, achieving super-additivity of binding energy.
• Fragment Merging: Combining structural features from two overlapping fragments into a single, more potent hybrid molecule. Structure-guided design using co-crystal structures is critical for all three approaches.

The fundamental advantage of FBS lies in its superior sampling of chemical diversity. A library of just 1,000 fragments can represent the same chemical diversity as a library of 1,000,000 drug-sized molecules. This is because the combinatorial explosion of possible molecules is constrained at the fragment level. FBS efficiently probes 'fragment space' to find core scaffolds, which are then decorated to explore 'lead-like space' in a rational, structure-guided manner.

A high-quality fragment library is the foundation of a successful FBS campaign. Key design principles include:

• Chemical Diversity: Maximizing the number of unique scaffolds and pharmacophores.
• Purity and Solubility: Ensuring fragments are highly soluble in aqueous buffer (typically > 1 mM in DMSO and assay buffer) to prevent aggregation and false positives.
• Rule-of-Three Compliance: Filtering for fragment-like physicochemical properties.
• Synthetic Tractability: Prioritizing fragments with vectors for straightforward chemical derivatization.
• Absence of PAINS: Rigorously filtering out known pan-assay interference compounds.

The strategic construction of a small (500–5,000 compounds) collection of low molecular weight (< 300 Da) fragments that obey the Rule of Three (Ro3). Key design principles include:

• High aqueous solubility (> 1 mM) for screening at high concentrations
• Chemical tractability with multiple synthetic vectors for fragment growth
• Absence of reactive or promiscuous functional groups
• High sp³ carbon fraction to maximize three-dimensionality and novelty
• Spectroscopic purity verified by NMR and LC-MS Libraries like the Maybridge Ro3 Fragment Library and Life Chemicals FBS Collection are commercially available starting points.

Unlike biochemical assays, fragment screening requires sensitive biophysical techniques to detect weak binding (KD ~ µM–mM). Core methods include:

• Surface Plasmon Resonance (SPR): Label-free detection of binding kinetics and affinity via changes in refractive index
• Nuclear Magnetic Resonance (NMR): Ligand-observed (STD, WaterLOGSY) and protein-observed (HSQC) experiments to map binding sites
• Thermal Shift Assay (TSA): Measures ligand-induced stabilization of protein thermal denaturation
• X-ray Crystallography: Direct visualization of fragment binding in the target's active site, enabling immediate structure-guided optimization
• Mass Spectrometry: Native MS and covalent labeling approaches for hit detection

Once fragment hits are validated, three primary strategies convert weak binders into high-affinity leads:

• Fragment Growing: Iterative addition of functional groups to explore adjacent binding pockets, guided by structural data and computational docking
• Fragment Linking: Connecting two fragments that bind to proximal but non-overlapping sites with a chemical linker, requiring careful optimization of linker length and geometry
• Fragment Merging: Combining structural features from two overlapping fragments into a single, more potent hybrid molecule
• Fragment Self-Assembly: Exploiting dynamic combinatorial chemistry where fragments reversibly link in the presence of the target protein, amplifying the highest-affinity combination

Fragment-based screening uses normalized potency metrics to compare hits of different sizes and prioritize optimization:

• Ligand Efficiency (LE): Binding free energy per heavy atom (ΔG / N_heavy), with values > 0.3 kcal/mol/atom considered promising
• Lipophilic Ligand Efficiency (LLE): pIC₅₀ – logP, penalizing potency derived from non-specific hydrophobicity; values > 5 are desirable
• Group Efficiency (GE): The contribution of a specific added functional group to binding affinity, guiding fragment growing decisions
• Fit Quality (FQ): A statistical measure comparing observed affinity to predicted affinity based on size, identifying fragments with optimal binding interactions
• Ligand Efficiency Dependent Lipophilicity (LELP): logP / LE, where values < 10 indicate favorable property profiles

In silico methods complement experimental FBS by pre-filtering libraries and predicting binding modes:

• Molecular Docking: High-throughput docking of fragment libraries into target structures, though scoring functions struggle with weak binders
• Hot Spot Mapping: Computational solvent mapping (e.g., FTMap, MDmix) identifies energetically favorable binding pockets on protein surfaces
• Pharmacophore-Based Screening: Defining minimal 3D feature requirements from known fragment hits to search larger virtual libraries
• Free Energy Calculations: FEP+ and MM-GBSA methods for accurate relative binding affinity predictions during fragment optimization
• Machine Learning Scoring: Deep learning models trained on fragment-protein co-crystal structures to improve hit ranking and false-positive rejection

FBS has produced multiple FDA-approved drugs, validating the approach:

• Vemurafenib (Zelboraf®): BRAF V600E inhibitor for melanoma, developed from a 238 µM fragment hit optimized to low nanomolar potency via structure-guided fragment growing
• Venetoclax (Venclexta®): BCL-2 inhibitor for chronic lymphocytic leukemia, originating from an NMR-based fragment screen and elaborated using fragment linking and growing
• Erdafitinib (Balversa®): FGFR inhibitor for urothelial carcinoma, discovered through fragment-based lead generation targeting the kinase hinge-binding region
• Pexidartinib (Turalio®): CSF1R inhibitor for tenosynovial giant cell tumor, developed from a fragment hit with initial LE > 0.4 kcal/mol/atom