Computational Approach for Combinatorial Library Design

1
Computational Approach for Combinatorial Library
Design

• Journal club-1
• Sushil Kumar Singh
• IBAB, Bangalore

2
Chemical Synthesis of compounds

• Traditional Synthesis
• A B ?AB(1 chemist50 compounds/Year)
• Combinatorial Synthesis
• A1.......Am
• B1.......Bn
• Total No of Compounds mn
• 1000 - 10,000 compounds per experiment.

3
Challenges

• Do these new technology will lead to better drugs
  faster ?
• Can we make and test every thing ?So,
• Lead discovery and optimization is not a just
  game of numbers and it requires intelligent
  design choice.

4
Library design method

• Setting up library with maximum diversity.
• Diverse libraries -
• Lead generation libraries for screening against
  no of targets
• A structurally diverse library should cover
  biological activity space as well.

5
Requirements for measuring diversity

• Molecular descriptors to define structure
  space 1D- mw, log p(o/w) 2D- surface area,
  flexibility. 3D-Pharmacophore- collections of
• atoms/functional group their
  orientations.
• Way to quantify the (dis)similarity of
  compounds.
• Subset selection algorithm to ensure full
  coverage of structural space.

6
Descriptors selection and validation

• On comparing 2-D and 3-D descriptors, it was
  found that 2-D descriptors were more effective
  and more accurate for structure search.
• 2-D descriptors consist majority of
  substructures present in the molecule.
• 3-D pharmacophore encodes information about
  only three atoms or group at a time with limited
  no of conformations but important for biological
  activity of molecule.

7
Descriptors and chemical space

• While choosing descriptors
• Avoid correlated descriptors.
• Choose those which can add maximum biological
  activity to library.
• Every descriptor adds a dimension to chemical
  space.
• A large no of descriptors is often reduce to
  smaller no using PCA.

8
Quantifications of (dis)similarity of compounds

• Tanimoto coefficient bc/(b1b2 bc)
• for 2-D molecular similarity by comparing
  bit strings. (e.g MDL information systems.)
• Fingerprints like Daylight, ISIS etc. also
  compare 2-D similarity.
• 3-D pharmacophore similarity also calculated by
  on the basis of bit string in 2D case.

9
Other methods

• Distance-based-
• MaxMin chooses points to maximize the smaller
  near-neighbor distance in design set.
• Grid/Cell based
• BCUT Combination of 2-D and 3-D
  descriptor commonly use in QSAR.
• 3-D pharmacophore maximizing diversity - rigid
  and flexible conformers with multiple energy.

10
Lead optimization

• After library generation ? Lead
  optimization
• -before optimizing a hit, do activity analysis of
  different regions of a molecule with small no.
  of individual molecule.

11
Conclusion

• Library design depends
• Design algorithm and
• Property space
• So Comparing library is a difficult different
  design with different property space.
• It requires
• Combination of structural diversity calculations.
• Experience.
• Good medicinal chemical intuition.

12

• Thank You