Title: Classical Electrostatics and Drug Design: Enumeration and Evaluation of Isosteric Analogs
1Classical Electrostatics and Drug Design
Enumeration and Evaluation of Isosteric Analogs
- Roger Sayle, Bob Tolbert, Andrew Grant and
Anthony Nicholls - OpenEye Scientific Software,
- Santa Fe, New Mexico
2Motivation Docking is Difficult
- Protein Flexibility
- Conformational Strain
- VdW Clash Resolution
- Entropic Effects
- Temperature Pressure
- Hydrogen Bonding
- Hydrophobic Effect
- Desolvation Terms
- Tautomerism
- pKa/Protonation State
3Cheating Docking Isosteres
- Isosteres are compounds with the same shape.
- For this talk, an isostere has the same heavy
atom topology and non-terminal hybridizaton state
(geometry). - Isosteres are more likely to adopt the same
binding mode than non-isosteric compounds. - Given a bound ligand-protein complex crystal
structure, we can extrapolate isosteres with more
confidence. - QSAR works well within compound series.
4Cheating is what were good at!
- All remaining docking terms are electrostatic.
- Use state-of-the-art partial charging methods and
continuum solvent electrostatics models. - Continuous dielectric Poisson-Boltmann.
- Modern bond-charge increment partial charges.
- Efficient and realistic isostere/state
generation. - Dramatic simplification in electrostatic
contribution evaluation. - WABE GIMBLE
5WABE Crosswords for chemists
6Solving the Jigsaw Puzzle
- Wabe reads a molecule database as a training set.
- An unique atom type forms a piece, consisting of
atomic number, formal charge, implicit hydrogen
count, heavy degree and valence. - The pieces edges are formed by the atom types of
its immediate neighbors. - A piece may be substituted into the puzzle if
its heavy degree and hybridization match the
template. - Different training sets can used to tailor the
acceptable results actives, non-toxic, known
chemistry, available.
7Example WABE solutions
8(No Transcript)
9Electrostatics in Water
10Electrostatics in an Active Site
11Perfect Charges
12Tidor Perfect Charges
13Gimble
- Using Tidors formulation of the matrix its
possible to calculate electrostatic binding
energies orders of magnitude faster than full
Poisson-Boltzmann calculations. - Full PB times (zap_bind)
- 1320 mols 54m44 2.48 s/mol
- Gimble times
- 1320 mols 1m34s ltlt 0.07 s/mol
14Gimble Approximation Accuracy
15Really CheatingKnowing the answer in advance
16Examples of Ring Substitutions
17The Best and the Worst
1199
- Better then methotrexate (221/1320)!
1047
18Specifics
Literature Charged WABE Monday Morning, 8am
19Specifics
Literature Doesnt Matter WABE Effect is small
20Specifics
Literature Essential WABE Essential
21Specifics
Literature Non-Essential WABE Non-Essential
22Specifics
Literature C better than N WABE Mais Oui!
23Learning from CheatingWhen you didnt really
know the correct answer.
Exptl IC50 3.1µM
Exptl IC50 0.125µM
- Hansch et al., J. Med. Chem., 20, 96-102, 1977.
- Piper et al., J. Med. Chem., 29, 1080-1086, 1986.
- Graffner-Nordberg et al., J. Med. Chem., 43,
3852-3861, 2000.
24But does it scale?
33,250 analogues, Gimble takes 2m41s, gt 200
mols/s
30609
26192
25Perhaps we were lucky?
Astra Zeneca M532121/p56 Lck Kinase
26X-ray Structure of Lck-M532121
1.5Å crystal complex (R19.7, Rfree21.7)
27We didnt know we could cheat
- Discovery of 2-Phenylamino-imidazo4,5-hisoquinol
in-9-ones A New Class of Inhibitors of Lck
Kinase, - R. Snow, M. Cardozo, T. Morwick, C. Busacca, Y.
Dong, R. Eckner, S. Jacober, S. Jakes, S.
Kapadia, S. Lukas, M. Panzenbeck, G. Peet, J.
Peterson, A. Prokopowicz, R. Sellati, R. Tolbert,
M. Tschantz and N. Moss. - J. Med Chem. 2002, 45, 3394.
28Experimental Assay Data
Ureas ? Thioureas lt Guanidines CH3/Cl lt CH3/CH3 ?
Cl/Cl lt F/F
29Gimble Scores
Ureas ? Thioureas lt Guanidines CH3/Cl lt CH3/CH3 ?
Cl/Cl lt F/F
301,792,000 compounds later...
Fluoro/Amino _at_2 Chloro/Methyl _at_3 Thiazinanium
_at_4 Methoxy _at_5 Oxazinanium _at_10 1,5-Naphthyridine
_at_13 Thiourea _at_15 Quinazoline _at_71 Urea _at_82
AZ Thiourea _at_360 AZ Urea _at_535
31The Timms Effect
Gimble (PB Electrostatics) picks up the critical
SAR of the solubilizing group thats pointing
into solvent.
32What worked and what didnt.
- MMFF94 charges perform significantly better than
Marsilli-Gasteiger charges. - Explicit hydrogens are important, but their
placement isnt critical. - Implicit hydrogens and united atom partial
charges are sufficient to identify formal charge
state, i.e. acid/base protonation, but not to
distinguish between functional groups.
33Improvements in Partial Charging
Similar technological advances in
electrostatics e.g. smoothed dielectric boundary
methods (Zap)
34Conclusions
- Residual problems with radii differences.
- Medicinal chemistry from first principles
physics. - The technical problems with docking may still be
with posing rather than scoring. - A need for different smooth posing functions
vs. accurate scoring functions?
35Acknowledgements
- Anthony, Bob and Geoff.
- Andrew Grant at AZ.
- Astra Zeneca.
- HP, IBM and SGI
- Lewis Carroll
36Bibliography
- Lewis Carroll, Through the Looking Glass.
- Lee-Peng Lee and Bruce Tidor, Optimization of
electrostatic binding free energy, J. Chem.
Phys., 106 (21), 8681-8690, June 1997. - Erik Kangas and Bruce Tidor, Electrostatic
Complementarity at Ligand Binding Sites
Application to Chorismate Mutase, J. Phys. Chem.
B., 105, 880-888, 2001. - C. Hansch, J.Y. Fukunaga, P.Y.C. Jow and J.B.
Hynes, Quantitative Structure-Activity
Relationships of Antimalarial and Dihydrofolate
Reductase Inhibition by Quinazolines and
5-Substituted Benzyl-2,4-diaminopyrimidines, J.
Med. Chem., 20 (1), 96-102, 1977. - M. Gaffer-Nordberg et al., Computational
Predictions of Binding Affinities to
Dihydrofolate Reductase Synthesis and Biological
Evaluation of Methotrexate Analogues, J. Med.
Chem., 43 (21), 3852-3861, 2000.