Fragment-Based Drug Discovery

• What makes a good drug?

  • Potent and selective effect on a given target
  • Biological effect (treat/cure the symptoms/cause of disease)
  • Appropriate set of physiochemical properties
  • Lipinski rule of five for orally active drugs
• The top 100 marketed drugs:

  • 18 bind to GPCR
  • 10 bind to nuclear receptors
  • 16 bind to ion channels
  • Most of remainder inhibit enzymes
  • 74% are chemical, 16% biological, 8% natural and 2% other
• Steps in developing a drug:

  • The hit must be optimised into a lead molecule, which is then optimised into a drug candidate
  • Optimisation relies on evaluation, synthesis and design
  • Paying attention to structure activity relationship and factors like solubility

  
• Methods of hit identification:

  • High throughput screening (HTS)
  • In silico approaches (structure-based, ligand-based)
  • Fragment-based drug discovery
  • Drug repurposing
  • Serendipity (e.g., penicillin)
• Fragments are small and more likely to bind to at least one protein pocket, increasing promiscuity. A linking strategy can be used to find a lead compound.
• Fragment hits differ from HTS hits, with Astex's rule of three for fragments.

  • MW ≤ 300 Da
  • ≤ 3 H-bond donors
  • ≤ 3 H-bond acceptors
  • CLogP ≤ 3
  • Rotatable bonds ≤ 3

  Most fragments have between 5 and 20 non-hydrogen atoms.

  
• Chemical space: an ensemble of combinations that can be made with a particular set of atoms.
• Effects of chemical space and fragments:

  • With four atom types (C, N, O and S) there are 10^63 compounds possible as 'drug-like' compounds
  • Largest libraries contain 10^8 compounds
  • A fragment has 10^8 possible compounds
  • A library of 1000 fragments ~ 0.01% of chemical space
  • This is more efficient than 'drug-like' molecules
• The use of chemical space translates into a higher hit rate than HTS, giving more starting points for optimisation.
• Ligand efficiency (LE): takes into account the number of non-hydrogen atoms which is reflective of the molecule size.
• Small fragments are less likely to have interfering functionality (molecular complexity).

  • Smaller libraries allow more up-front attention to purity and drug-like properties
  • Smaller libraries are easier for universities and small companies to get started
  • Fragments can tackle new classes of targets

  
•  When designing libraries:

  • Can be assembled in house
  • Can be purchased
  • Can be implemented with in-house fragments that are underrepresented
  • Consideration must be given to composition and size
  • Must be diverse as possible
  • Must be stable, without reactive/toxic moieties
  • Solubility up to 100 mM in DMSO, > 1 mM in aq. buffer
  • Should be synthetically tractable

  
• Library design

  • Usually rich in 2D features (aromatic and heteroaromatics)
  • Conformationally constrained, provides rigidity (entropy)
  • Easily identifiable by ligand observed NMR
  • Can be biased depending on targets e.g., kinases, metalloenzymes, protein-protein interactions
  • Can be biased based on screening technique
  • High solubility for X-ray crystallography
  • Solubilised in d6-DMSO for NMR studies
  • Containing fluorine atoms for 19F NMR
  • Can contain fragments derived from FDA approved drugs
  • 3D fragments

  
• Fragments are inherently weak binders (KD = 0.1 - 10 mM).

  • Sensitive and robust methods for detecting weak interactions are essential
  • Biophysical techniques favoured over biochemical techniques
  • More sensitive, typically produce fewer false positives
• Surface Plasmon Resonance (SPR)

  • Changes in refractive index near surface of sensor chip dependent on mass of surface layer
  • Direct binding: target molecule fixed to chip
  • Binding causes changes to the refractive index (magnitude proportional to increase in surface mass)
  • Binding affinity can be calculated (KD = 1/KA = kdissoc/kassoc)
  • Low protein consumption, high throughput
  • Usually used as a primary screen, can detect fragments ≥ 100 g/mol

  
• Differential Scanning Fluorimetry (DSF) - the thermal shift assay

  • Thermal protein unfolding monitored using environmentally sensitive fluorescent dye
  • Ligands increase unfolding temperature of target by binding to and stabilising folded state
  • Typically forms the basis of a primary screen, low protein consumption (low μM), 96/364 wp, 20 μL per well, 102 - 103 fragments per day
  • Only requires a PCR machine
  • Tm influenced by the experimental conditions (buffer, salts, detergents, reducing agents)
  • Prone to false positives
  • Does not provide accurate affinity constant (Kd)

  
• False positives can be introduced by fluorescent or reactive compounds, or if they interfere with the dye in DSF.
• Ligand Observed NMR

  • Affects the NMR signals of the small molecule ligand
  • No labelling required
  • No size limit for the protein
  • Medium protein consumption (micro M)
  • Rapid data acquisition
  • Semi quantitative information
  • No or few structural information e.g., binding mode
  • Suitable for weak binders with fast exchange rates
  • Small molecules can be screened in cocktails to increase throughput (up to 5)

  
• Protein observed NMR:

  • Affects the NMR signals of the protein
  • Requires labelling of the protein (15N)
  • Suitable for proteins <~ 40 kDa
  • High protein concentrations required (50 micro M - mM)
  • Slow data acquisition
  • Provides structural information
  • Suitable for weak and strong binders
  • Low throughput (1 ligand at a time) with several samples required to determine KD

  
• Isothermal titration calorimetry (ITC)

  • Access the thermodynamic signatures of binding
  • Access to Kd and its components ΔG, ΔH and ΔS (thermodynamic signature)
  • Low throughput, high protein consumption (> 30 μM)
  • Used for characterisation of the hits

  
• Choosing the right fragment hit to follow-up

  • Filtering: chemical stability and PAINS? Remove unstable/reactive compounds and PAIN substructures. In theory these should have been discarded upstream, during the design of the library. Known inherent toxicity of the scaffold?
  • Prioritise: Potency (Kd low mM or better), solubility (low mM in buffer), LE (> 0.3), synthetic tractability (analogues must be rapidly available through established chemical reactions)
• Fragment elaboration II: Merging

  • Possible when two or more fragments bind in overlapping regions of binding pocket
  • Fragments are merged to increase affinity

  
• Fragment elaboration III: Linking

  • Possible when 2 or more fragments bind in adjacent but non-overlapping regions
  • A linker that retains the position and orientation of the fragments is designed
  • Structural information is required to guide linking

  
• Fragment elaboration I: Growing

  • Most common method of fragment elaboration requires only a single fragment hit
  • Fragment is grown step-wise to pick up additional interactions
  • Structural information on how the fragment binds to the target is essential

  
• Computer aided drug design/discovery:

  • Semi quantitative predictions of protein-ligand interaction/affinity
  • Hit-to-lead or lead optimisation
  • Trade-off in terms of throughput, computing requirements, time consumption, accuracy, prior knowledge
• Computer aided drug design/discovery methods:

  • In practice, computational models are refined with experimental data and vice versa
  • Docking
  • Molecular dynamics
  • QM based methods

  
• Structural Issues I:

  • Bromide is easily displaced by SN2 by any nucleophile on a protein (curly arrows mechanism with cysteine)

  
• Structural Issues II:

  • Sulphonate chlorides are good electrophiles
  • Will be displaced by SN2 by any nucleophile on a protein e.g., cysteine

  
• Structural Issues III:

  • Hydrazones are more stable than imine (will not hydrolyse in water)
  • Hydroxyl group position will cause the molecule to chelate with metals (EWG will make it act like a carbonyl, easy O- formed)

  
• Structural Issues IV:

  • Trifluoromethylester - good EWG with inductive effects, which will make the carbonyl more delta positive than a typical ester
  • Reactions such as transesterification

  
• Structural Issues V:

  • The more electron donating, the more likely the reaction of the hydroxyl
  • Amine will protonate easily, which can be lost as a leaving group, making an intermediate which is a quinone methide (a Michael acceptor)
  • Nucleophilic groups on proteins will then react
  • Very toxic and can act as metal chelators

  
• Structural Issues VI:

  • Catechols will chelate metals

  
• Structural Issues VII:

  • Azo dyes absorb light (UV-Vis) and are brightly coloured
  • They can isomerise with light from trans to cis, causing it to be a false positive

  
• Structural Issues VIII:

  • Michael acceptor
  • Reacts easily with cysteine
  • Reacts with nucleophilic groups

  
• Structural Issues IX:

  • The more electron donating, the more likely the reaction of the hydroxyl
  • Benzyl ether will protonate easily, which can be lost as a leaving group
  • Nucleophilic groups on proteins will then react
  • Very toxic and can act as metal chelators

  
• Structural Issues X:

  • Strong Michael acceptor
  • Will react with cysteine at neutral pH
  • Will react with nucleophiles

  
• Structural Issues XI:

  • Epoxide is strong Michael acceptor
  • Will react with nucleophiles
  • Reacts with cysteine

  
• Structural Issues XII:

  • DNA intercalator
  • Generates defects leading to mutations

  
• Notes on drug structures:

  • Epoxides should be avoided due to electrophilicity
  • Michael acceptors should be avoided due to nucleophilicity and potential for conjugate addition
  • Can be used to label a protein with a particular functional group
  • Not a good idea for a drug as it will react with many different compounds/cells
  • Scaffolds that are known for intrinsic activity e.g., dioxenes
  • Synthetic tractability is important - control of stereogenic centres (the more stereogenic centres, the harder the synthesis and establishing structure-relationship)
• Enzyme catalysis

  • Enzymes are catalysts that accelerate reactions
  • They do this by lowering the activation energy
  • Rate constant (k) related to activation energy by the Arrhenius equation