Drug discovery

• The general process for identifying a novel drug:

  • Relevant mechanism identified by structure-function studies (NMR, x-ray crystallography, computation)
  • Mechanism-based screens if pharmacophore known OR high-throughput screens if pharmacophore not known (wet lab)
  • Lead compound established (best drug based on screening)
  • Drug candidate proposed for preclinical development and clinical testing
• Using high-throughput screening can be disadvantageous because it is expensive, creates false positives and there may be errors in the data
• The two types of computer-aided drug discovery are ligand-based methods (LBDD) and structure-based methods (SBDD)
• SBDD is used when the 3D structure of the target protein is known
• LBDD is used when the 3D structure of the target protein is not known; exploits information about the molecules that are known to bind to the target protein to determine what will bind to the target protein best
• LBDD strategies take into account molecular information such as 3D molecular shape, molecular and electronic properties and 3D pharmacophore
• The pharmacophore is the spatial arrangement of chemical features responsible for biological activity - i.e. the part of a target protein that will bind to a drug/inhibitor
• LBDD is kind of like a positive feedback loop: with every library search, the pharmacophore generated becomes more accurate, so can be used to narrow the library search.
• LBDD via similarity measure involves combining molecular descriptors and the similarity coefficient to rank a database of compounds on similarity to the reference compound, to identify the best ligand for the target protein
• Physicochemical properties used as part of LBDD similarity measure include molecular weight, hydrophobicity score (logP), no. of atoms (Mr)
• 2D properties used in similarity measures for LBDD include fingerprints (to identify functional groups), topological indices (how the functional groups are linked to each other) and maximum common substructures (how similar the compound needs to be to the exemplar compound)
• A 2D fingerprint is a bit string where each bit reflects the presence (1) or absence (0) of each fragment in the molecule. Similarity is based on determining the number of bits that are common to two structures.
• The hashed fingerprint is a bit string that contains information on the topology - the proximity of different functional groups
• Bit collision in a hashed fingerprint refers to when the same bit is set by multiple patterns. Too many bit collisions in a single fingerprint may result in the loss of information
• The Tanimoto coefficient is used to determine the similarity between the parent compound and the library compounds
• Tanimoto coefficient:

  T(a,b) = Nc / (Na + Nb - Nc) where:

  • c bits set in common in the reference and database structure
  • a bits set in reference structure
  • b bits set in database structure
• Scaffold hopping is used to find a molecule that has the same activity as the query molecules, but with a different core structure
• Scaffold hopping can be broken down into:

  • Shape matching - i.e. matching the shape of the electron cloud
  • Pharmacophore searching
  • Fragment replacement - modification of the structure to improve the binding properties
  • Similarity search
• 3D fingerprinting encodes the 3D features into the bit string according to the distance between relevant functional groups
• 3D fingerprinting exploits:

  • Pairs of atoms at given distance range
  • Triplets of atoms and associated distance
  • Pharmacophore pairs and triplets e.g. donors, acceptors, aromatic centres
  • Valence angles
• When scoring compounds based on their structures, include decoy compounds that are known to be unable to bind to the protein; these should not appear on the ranked list, so if they do, you know you have made an error