Fast Computational Method for Fragment Growing and Joining Using Molecular Fields

Dr Martin J Slater

Cresset BMD: Who are we?

> Cresset was founded in 2002 by innovator Andy Vinter > We provide and continually develop a suite of unique cutting

edge molecular modelling software for drug discovery.> We have clients from big pharma, agrochem, biotech and

academia

Cresset’s unique technology

> It uses a condensed 3D representation of the electrostatic, hydrophobic and shape properties of molecules together with the full fields.

3D Molecular Electrostatic Potential (MEP)

What proteins see?

Field Points2D Representation

What chemists see?= Positive = Negative

= Shape= Hydrophobic

What our CPUs see?

> Field patterns from Cresset’s proprietary XED force field reproduce experimental results

XEDs make fields work

Interaction of Acetone and Any-OH from small molecule

crystal structures

Experimental Using XEDs

C O

-0.5

-0.5

-0.5

-0.5

-0.5

-1.75

-1.75

+5

+1

H

-0.5

-0.5

+0.9+0.1

Not using XEDs

XED adds p-orbitals to get better representation of atoms

Biologically Relevant Molecular Comparisons

Bioisosteres

Bioisosteric groups

View fields

N FNO

Field points give you new insights into your molecule

Experimental

(Data from small molecule xray

structures)

Fields

Structure

Comparing 2D and 3D metrics

Similar 2D=3D_FS

2D=3DDissimilar 2D=3D_FS

2D = 3D_FS

Example - Higher 3D Sim

2D sim = 0.1(other methods=0.3)

SO Br

N

HNO

3D field sim = 0.82

O

O

Example - Higher 3D Sim

141

2D sim = 0.2

3D sim = 0.7

454

O

O

HN

Spark’s Approach

Rofecoxib

Valdecoxib

Etoricoxib

12nM

> Find bioisosteres by replacing sections of the molecule

Spark’s Approach

1 Select a region to replace and remove these atoms

Wrong distance

Spark’s Approach

1 Select a region to replace and remove these atoms2 Search database for matching fragments

(geometric search only)(search runs on fragment conformations)

Spark’s Approach

Wrong angle

1 Select a region to replace and remove these atoms2 Search database for matching fragments

(geometric search only)(search runs on fragment conformations)

1 Select a region to replace and remove these atoms2 Search database for matching fragments

• geometric search only(search runs on fragment conformations)

3 Form Products• minimise and add field points

Spark’s Approach

Good match

1 Select a region to replace and remove these atoms2 Search database for matching fragments

• geometric search only• search runs on fragment conformations

3 Form Products• minimise and add field points

4 Score

Spark’s Approach

0.88

> Produces more diverse, non-obvious bioisosteres> Avoids fragment scoring

limitations> Allows for electronic influence of

replacing a moiety on the rest of the molecule and vice versa

> Allows for neighbouring group effects

Whole-Molecule Scoring Advantages

OO

S OO

SN

S OO

NN

Example - COX-2

> Search for Bioisosteres for cyclic lactone of Rofecoxib

O O

SOO

aromatic atoms only

Actives:9 of the first 10 clusters21 of the first 30 clusters

Search Common Dbs

87,225 frags

Cluster, Result Nos

Cluster id

FieldStere result 2D sim closest lit compd

10,12 0.52 Same, 60nM

13,19 0.509

7nM

32, 104 0.48 None

COX-2 Results

Cluster, Result Nos

Cluster id SPARK result 2D sim closest lit

compd

1,1 0.646

2,2 0.531

6.5uM

5,5 0.500 Same, 10nM

9,11 0.532

70nM

ON

SO

O

ON

SO

OH2N

ON

NNS

SO

O

NNS

SO

O

NNS

F

NS

NN

SO

O

NS

NN

S

SO

O

S

O NH2

SO

OH2N

S

O O

N

SO

O

N

SO

O

F

F

SO

O

F

F

F

N NNN

SO

O

N NNN

Scaffold replacement ‘Sildanafil’

spark (10 mins)NEAT?

spark results 2-5

Fsim score 0.985 0.985 0.982 0.978

spark results 6, 7, 10, 13

Fsim score 0.973 0.973 0.959 0.957

spark results 14, 18, 26, 30

Fsim score 0.953 0.951 0.943 0.939

HN

NO

O

HN

N

NN

O

HN

N NN

O

HN

NN

N

O

S

NN

N

O

O

NN

N

O

HN

N N

N

O

O NN

HN

N

O

ON

N

O

HN

N

N

O

HN

NN

O

N

NHN

N

O

O

Pfizer J. Chem. Inf. Model. 2012

Fragment growing example

> FieldStere version 3.0.0 fragment growth example:

1. P38 kinase bound to a fragment fluorescent probe PDB:3K3I specific to the ‘DFG-out’ conformation

2. ‘DFG-in’ example with specificity towards the ‘Gly’ flipped hinge PDB:3ROC and/or 3HUB

> Selectivity potentially to be gained by combining ‘Gly flip’ and ‘DFG-out’ in one molecule

> Can we use the new version of SparkV10 to grow the DFG-out fragment into the DFG-in hinge?

Fragment in DFG-out pocket, PDB:3K3I

predominant hingeconformation

Graphics from Pymol from Delano Scientific

+Gly hinge flip ligand_1, PDB:3ROC

Hinge Gly flip

Graphics from Pymol from Delano Scientific

Fragment compatibility – DFG-out

Structure Rank Sim Structure Rank Sim

1 0.528 11 0.492

4 0.507 13 0.489

6 0.501 53 0.476

Fieldstere output: 2D mols

Fieldstere output: 3D mols and fields

Fragment and reference Rank 4 Rank 6

Rank 11 Rank 13 Rank 53

Outcome

> Fragment growth both possible and a facile using an automated process with SparkV10

> Interesting and sensible candidate molecules generated

> Predict highly selective p38 actives

> Absolute requirement for 3D insight

Any relevance? – Pfizer compound for COPD

Deposited in PDB: 2YIS November 2011

Clinical trials for COPD

Virtual compounds in the output list……prepared for ACS San Diego Late Summer 2011

S

OH

N

NN

S

NH

ONH

NN OH

F

N

NN

NHN

Br

N

S

O

SH2N H

N

N

Br

O

NH2

O

N

Conclusion

> Cresset offer a wide variety of software and collaborative solutions for drug discovery

> Cutting edge technology

> Provide key insights

martin@cresset-group.comQuestions welcomed