; An Ab Initio Fragment Database for use in Drug Design
Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

An Ab Initio Fragment Database for use in Drug Design

VIEWS: 12 PAGES: 27

  • pg 1
									An Ab Initio Fragment Database for
  use in Predicting Bioisosteric
  Replacements in Drug Design

             Michael Devereux (University of Manchester)
             Dr PLA Popelier (University of Manchester)
Introduction
   Produced the ‘Quantum Isostere Database’, a                              O
    web-based tool to aid medicinal chemists in                                      O
    lead optimization.                                                   S
                                                         O       S

   Addresses issue of ‘bioisosterism’ in                    N
                                                             H
                                                                     N

    chemical fragments.
                                                                                     N
                                                                                         O
   Suggests novel, drug-like replacements for
                                                                                 S
    fragments in lead compounds.
                                                             O       S

                                                                 N           N
   Uses ab initio data for each fragment, variety               H
    of descriptors from partitioned charge              Example taken from Lima,
    density.
                                                     Barreiro, Curr. Med. Chem. 2005
Quantum Chemical Topology (QCT)
   Atom, bond and fragment properties
    generated using ‘QCT’.

   Charge density partitioned, properties
    integrated over bounded regions.

   Bond properties derived from electron
    density at saddle points between
    nuclei.

   Properties additive – obtain fragment    Selected QCT atoms in Prozac molecule.
    properties by combining atomic           Two different isodensity surfaces are shown.
    contributions.
The Quantum Isostere Database (QID)

   Web-based to allow global
    access to database.

   Users sketch in fragments and
    identify key characteristics as
    search criteria.

   Provides fast access to
    chemical information,
    complements knowledge and
    experience of medicinal           Search options for basic fragment search. Can be
    chemists.                         combined with individual atom and bond criteria.
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
    Fragment Generation
   Linkers obtained by SMARTS
    fragmentation of the World Drug Index
                                                                                          N
                                                                  H
    (WDI) – improves ‘drug likeness’.                             N                               N
                                                                                  S
                                                                      N                       N       N
                                                                                              H       H
   Only fragments with 2 connection points to
    either a carbon or nitrogen atom on each          H
                                                                                                                              N
                                                      N
    side were considered.                                     *                                                       N
                                                                          S                           H
                                                  N               *           *                       N       *
                                                                                      S       *                   N       *
                                                          *                                       *       *       H
                                                                                  *
   Fragments can contain ring systems, but not
    form part of a larger ring.
                                                          Fragmentation of Cimetidine molecule
                                                              showing typical ‘linker’ groups.
   2597 two-atom linkers, 982 one-atom
    linkers were found.
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
Capping Groups
                                                                                H
                                                        H                       N         O
                                                        N     O                                      H
                                                                                                     N         O
                                                                       Et            Et
   Investigated transferability of            Et            Py
                                                                                               Ph         Et
    QCT properties.                        H
                                           N         O                                                         H
                                                                                                               N       O
                                                                            H
                                                                            N        O
   3 different capping groups        Py        Et
                                                                                                         Ph         Ph
    represent most carbon and
                                           H                       *             *
    nitrogen environments.                 N        O                                                     H
                                                                                                          N        O

                                      Py       Ph             H
   Focus on ethyl environment in                             N    O                 H              Ph         Py
                                                                                     N        O
    current work.
                                                        Py        Py
                                                                            Et            Ph                   Py=Pyrrole

                                                         Different permutations of capping
                                                               groups for each linker
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
Conformational Search




                                          N
                                  N

                                      N
                                      H

   Systematic conformational searches carried out in Macromodel. Only
   bonds inside the linker group (not the capping group) were rotated. 20
   most varied conformers kept for each linker.
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
Properties Available
   Steric properties: conformation, size, shape


   Electronic properties: multipole moments,
    electrostatic potential, kinetic energy and other
    properties derived from the density.


   H-bonding: Hydrogen bond acidity and basicity
    from properties of charge density.


   Bond properties: bond order and properties
    derived from the charge density at a “bond
    critical point” (bcp)
                                                        Acrolein molecule showing QCT
   Fingerprints: describe distribution of properties   atomic basins, atomic interaction
    in 3D space using grid.                               lines and bond critical points.
Ab Initio Surface Properties




  Local Ionization Energy (P.                                     Electrostatic Potential on
  Politzer et al.) on THF 0.001                                   0.001au isodensity surface of
  au isodensity surface                                           THF molecule




                                  Electrostatic Potential on
                                  0.001au isodensity surface of
                                  MeCN cyano group
H-Bonding Descriptors: α                                                            H
                                                                                    2


                                                                            1.000

                                                                            0.900

                                                                            0.800
                                    2
                                   R = 0.9208                               0.700

                                                                            0.600

                                                                            0.500
                                                                                                n=38
                                                                            0.400             r2=0.921
                                                                            0.300

                                                                            0.200

                                                                            0.100

                                                                             0.000
        -1.08   -1.06   -1.04   -1.02    -1     -0.98   -0.96   -0.94   -0.92      -0.9



        E.P.(Nuc) vs. H-bond acidity (c.f. Platts et al.)

 Observed vs. predicted H-bond acidity: α
                                                        H
                                                        2       N.B. Not:       ∑    α    H
                                                                                          2
H-Bonding Descriptors: β                                  H
                                                          2




                                                                     H-Bond bases - no
                                                                         N-atoms

                                                                           n=25
                                                                         r2=0.917
                                                                         q2=0.816




Observed vs. predicted H-bond basicity: β 2
                                          H
                                              N.B. Not:   ∑   β 2H
H-Bonding Descriptors: β                                                                                  H
                                                                                                          2


                                  betat2h_resiv.M1 (PLS), Untitled, PS-betat2h_resiv
                                  YPredPS[Comp. 1](YVar beta2h)/YVarPS(YVar beta2h)



                    0.50

                                                                                                                 H-Bond bases - no
                    0.40
                                                                                                                     N-atoms
   YVarPS(beta2h)




                    0.30
                                                                                                                       n=23
                    0.20                                                                                             r2=0.869
                    0.10
                                                                                                                     q2=0.837

                           0.10              0.20                    0.30              0.40        0.50
                                                       YPredPS[1](beta2h)



                                                    RMSEP = ---




Observed vs. predicted H-bond basicity: β 2
                                                                             H
                                                                                       N.B. Not:     ∑    β 2H
Geometric Comparison
   ‘Linker bonds’ joining fragment to
    capping group are used for alignment.

                                               *   *
   First bond is laid along the z-axis, 2nd
    capping group atom is rotated into
    YZ plane.

   Spatial coordinates of atoms can then
    be compared for all linkers.

   Conformation summarized using
    selected angles and distances of
    atoms in the 2nd linker bond.
3D Fingerprints
   Spatial distribution of atoms
    summarized using fingerprints -
    allows easy similarity searching.

   Each atom occupies a number of 1Å
    elements in a 22 Å x 22 Å x 19 Å 3D
    grid. Approach similar to GAP
    method of Leach et al.

   Tanimoto scoring used to compare
    shape similarity of fragments.

   Low occupancy of fingerprint makes    00000010000110000000001100000
    approach less efficient than
    traditional fingerprint methods.
3D Fingerprints
   Can summarize spatial distribution of
    any property in this way.



   Currently include positions of
    positive, negative and neutral atoms
    as separate ‘charge fingerprints’.




                                            00000010000110000000001100000
Database Construction
   1) Fragmentation of WDI.
    2597 two-atom linkers,           2) Filter Fragments
     982 one-atom linkers.                  (2360)
                                                                  3) Addition of
                                                               Chemical Environment


                                4) Macromodel Conformational
                                   Search on Linkers (~26K)


  5) Ab Initio Optimization
 800 high priority linkers so
              far.                   6) Properties Derived
                                     From Charge Density
                                                               7) Entry into Database
QID in use – 1
Find replacement with similar shape and conformation that has an H-bond acceptor of similar
                              strength at the Oxygen position.

                                O
                                        2
                        1                   *
  Query: amide linker       *       N
                                    H




 Search on Trans amide                           Search on Cis amide
QID in use – 2
                                     Search                                                 Criteria           Hits   Rank
                             O                               O
                                         O                                           Shape, Conformation,
    1                1           S                           S                                                  3      1
                         *       2               1
                                                     *
                                                                     2
                                                                     *               Charge Distribution
                                     *

                                                                     H
                             O                               O                       Shape, Conformation,
    2                                                                                                           16     1
                 1                   2                                               Charge Distribution
                                                                                                                               Database search results
                                                     1               2
                     *               *                   *               *


    3        1               O       2           1                   2               Shape, Conformation        7      1      compared with common
                 *                       *           *                   *

                     1                               1
                                                                                                                               chemical replacements
                         *                               *
    4
                                         2
                                             *
                                                             N
                                                                     2
                                                                         *
                                                                                     Shape, Conformation        40     2       identified from MDDR
                     1
                         *
                                                     1
                                                         *
                                                                                                                             database (Sheridan, JCICS,
                                                                                     Shape, Conformation,
    5
                                         2
                                             *
                                                             N
                                                                     2
                                                                         *
                                                                                     Bond Order
                                                                                                                5      1
                                                                                                                                        2001)
                         O                                       H           2
                                         2           1           N               *   Shape, Conformation,
    6        1                               *           *                                                      14     1
                 *               N
                                 H
                                                                                     Charge Polarity
                                                                     O


                             S                               O
    7            1
                     *
                                     2
                                         *
                                                 1
                                                     *
                                                                         2           Shape, Conformation        7      6
                                                                         *

                                                             O
                             S
    8            1
                     *
                                     2
                                         *                   S                       Shape, Conformation        7      1
                                                 1                   2
                                                 *                       *
                         O                                   O
                                         2                                           Shape, Conformation,
    9    1                                   *
                                                                             2
                                                                                 *                              23     1
             *                   N               1
                                                     *               O               Charge Distribution
                                 H
               O O                                               O                   Frag Dipole, Charge
                   2                                                         2
    10   1     S     *                               1                           *   Polarity, Electrostatic    12     1
             *   N                                       *           N
                 H                                                   H               Potential on Surface
QID in use – 3

Fingerprint Search




NC(=O)OC([2*])C([1*])C=CC=C




                              Fingerprint search to find closest shape matches –
                                         ranked by Tanimoto score
Summary

   We have exploited novel ‘QCT’ descriptors to characterise organic
    fragments.

   The descriptors allowed direct comparison of structural / electronic
    properties of different atoms, bonds and fragments.

   Data has been stored in the ‘QID’ database for easy searching and retrieval
    of bioisosteric replacements for linker groups.

   It is hoped this will become a standard tool to assist the Medicinal Chemist.
Acknowledgements
My sincere thanks to all those at Computational Chemistry -
GlaxoSmithKline (Stevenage), and Colleagues from the University of
Manchester.
In particular:


Michael Kranz (GSK)
Olivier Lamarche (Formerly University of Manchester)
Xiao Lewell (GSK)
Iain McLay (GSK)
Dan Ormsby (Accelrys)
Paul Popelier (University of Manchester)
Charlotte Quail (University of Manchester)
Michel Rafat (University of Manchester)

								
To top