Insilico screening without structural comparisons: Peptides to nonpeptides in one step Maybridge Wor

1
In-silico screening without structural
comparisonsPeptides to non-peptidesin one
stepMaybridge Workshop 23-24 Oct 03Bregenz
Austria
2
Cresset Biomolecular Discovery
Founded in November 2001 Funding by The Wellcome
Trust
3
Virtual Screening

• Virtual screening is the process of trying to
  find biologically-active molecules using a
  computer
• Protein-based (X-ray, docking)
• Need a protein structure
• Problems with scoring functions
• Ligand-based
• Structural similarity
• Not specific enough

4
The Science Problem

• The Problem is that
• There is no logical way to change Structural
  Class and retain Biological Activity
• Since we know that
• Different structures can give the same biological
  effect
• Then the Answer is to
• Define what it is that the target actually sees
  if not structure

5
Fields, XEDs and FieldPrints

• Fields
• A new method of describing molecular properties
• XEDs
• A new molecular modelling approach
• FieldPrints
• A new virtual screening method

6
Fields

• Chemically different, biologically similar
  molecules have a similar electron cloud.It is
  this that is seen by the target
• Can we use a representation of that electron
  cloud to explore molecules biological
  properties?
• Fields represent the key binding information
  contained in the electron cloud

7
COX-2 Inhibitor
8
COX-2 Inhibitor
9
COX-2 Inhibitor
10
COX-2 Inhibitor
11
COX-2 Inhibitor
R. P. Apaya, B. Lucchese, S. L. Price and J. G.
Vinter, (1995), J. Comp-Aid. Mol. Design, 9,
33-43.
12
The Field Template for a COX-2 Inhibitor
13
ACCs get Fields Wrong

• Without a good description of atoms, the field
  points are incorrect!

Atom-centred charges
Fields from ACCs
R. P. Apaya, B. Lucchese, S. L. Price and J. G.
Vinter, (1995), The matching of electrostatic
extrema A useful method in drug design? A study
of phosphodiesterase III inhibitors, J.
Comp-Aid. Mol. Design, 9, 33-43.
14
XEDs make Fields work

• The Field Points from XED agree well with those
  obtained from Quantum Mechanics

ACCs
XEDs
Vinter Trollope 1994 unpublished.
15
eXtended Electron Distributions
The XED force field improves the description of
electrostatics by extending electrons away from
the nucleus
XEDs
ACCs
J. G. Vinter, (1994) Extended electron
distributions applied to the molecular mechanics
of intermolecular interactions, J Comp-Aid Mol
Design, 8, 653-668.
16
XEDs Model Life Better
X-ray structure of Benzene
Benzene docked onto Benzene using XEDs
Benzene docked onto Benzene using ACCs
17
Aromatic-Aromatic Interactions
GSK (SKF) Azepanone-Based Inhibitors of Human
and Rat Cathepsin K, J. Med. Chem. 2001, Vol.
44, No. 9
18
Aromatic-Aromatic Interactions
19
XEDs - Summary

• A much better treatment of electrostatics
• Simplified force field
• Hydrogen bonding
• Anomeric and gauche effects
• Aromatic-aromatic interactions

20
Fields direct ligand binding mode
Dihydrofolate Reductase




21
Fields - Summary

• Proteins eye view
• Represent electron cloud NOT structure
• Distillate of important binding information
• Peptide/Steroid/Organic treated identically

J. G. Vinter and K. I. Trollope, (1995).
Multi-conformational Composite Molecular Fields
in the Analysis of Drug Design. Methodology and
First Evaluation using 5HT and Histamine Action
as examples, J. Comp-Aid. Mol Design, 9, 297-307.
22
Virtual Screening with Fields

• If field points are describing the binding
  properties of molecules
• Can they be used for virtual screening?
• Can we construct a fast accurate way of
  searching a Field Database?

23
FieldPrint Search Method
0010100100101
24
FieldPrintTM Limitations

• Fields are conformation dependent
• Need to populate database with conformations, not
  molecules
• Need to search with a specific conformation
• Throwing away some information (eg chirality)

25
Conformation Search

• Pop 2D to 3D
• Twist bond search
• Minimisation of all found conformations
• Filtered using 1.5Å RMSD
• 6 Kcal mol-1 cut-off
• Keep 50 conformations
• Rings not flexed amides forced trans

26
The Database

• The current database contains 2,500,000
  commercially available compounds
• 50 conformations stored for each compound
  (125,000,000 conformations)
• Results consist of similarity score for whole
  database
• Hits can be filtered (e.g. supplier, MW, Lipinski
  etc.)

27
Refinement

• The FieldPrint search front-loads the database
• We refine the FieldPrint results by performing
  true 3D field overlays
• Overlays are usually performed on the top 10-20
  of the database (ranked by FieldPrint score)
• Results are expressed as a field similarity

28
The 3D Field Overlay Principle
29
Fields Examples
PEPTIDE to NON-PEPTIDE
30
FieldPrint Performance
Retrieval of known inhibitors (spikes) from
600,000 compounds
Thrombin (49 Spikes) PPACK (D-Phe-Pro-Arg-CH2Cl)
spikes found
ranked database screened
31
FieldPrint - Thrombin Spikes
32
FieldPrint Performance (2)
Retrieval of known inhibitors (spikes) from
600,000 compounds
HIV NNRTI (52 Spikes)

• COX-2 Inhibitors (32 Spikes)

33
Validation

• James Black Foundation (JBF) funded by
  JohnsonJohnson
• GPCR target
• Exhausted Medicinal Chemistry of current series.
  Molecule in clinical development
• Back-up series required
• Two active diverse molecules available for
  template
• 3 Month deadline
• Commission mid-August 2002.
• Generate and search database. Supply list of
  compounds by mid-October 2002.
• Results returned early December 2002

34
FieldPrint Validation
Collaboration with the James Black Foundation
Distilled to 1000 Compounds
A GPCR (43 Spikes)
Visual inspection to 100
88 Purchased and tested
27 had pKb gt 5 (better than 10mM)
4 had pKb gt 6 (better than 1mM)
No structural similarity to any known actives. MW
range 350-600
35
Intelligent Lead Discovery

• Change structural class e.g. peptides to
  non-peptides, steroids to non-steroids
• As well as proteases, kinases (X-ray information)
• we can
• handle poorly defined targets e.g. GPCRs, Ion
  Channels
• because
• no protein data is necessary
• and
• minimal ligand 2D data is required

36
Where can Cresset be used?
Fast and flexible lead finding for new programs
allowing multiple starting points for medicinal
chemistry programs Lead switching on existing
programs Patent busting Moving away from ADMET
problems Finding back up series
37
Why should Cresset be used?
Diverse Structural Classes with Same Function
B
A
Peptide to non-peptide
Cost in Time and Money
Significantly faster than conventional
routes Cresset could go from A to B in
weeks Merck took 3 (?) years with 10 (?)
Medicinal Chemists!
Much more cost effective than HTS HTS 2,500,000
molecules _at_ 1 per molecule Cresset distils
this to just a few hundred!
38
Acknowledgements

• Cresset
• Dr J. G. Vinter
• Dr T. J. Cheeseright
• Dr M. D. Mackey
• Dr Sally Rose (consultant)
• James Black Foundation (KCL, JnJ sponsored)
• Prof. C. Hunter (Sheffield University)
• The Wellcome Trust

39
Intelligent Lead Discovery