Automated Estimation of Hammett Constants - … · • Study the impact of substitutions on electronic structure • Only relevant and effective among series of congeners – Topological,

SpectroscopySpectroscopy •• ChromatographyChromatography •• PhysChemPhysChem •• NamingNaming

Drawing and Drawing and Databasing Databasing •• Enterprise SolutionsEnterprise Solutions

Automated Estimation Automated Estimation of Hammett Constantsof Hammett Constants

Robert DeWitte & Ed KolovanovAdvanced Chemistry Development

Integrated Integrated

Software Software

SolutionsSolutions SpectroscopySpectroscopy •• ChromatographyChromatography •• PhysChemPhysChem •• NamingNaming


Let’s go back to our roots

Hammett, Taft, Hansch, Leo– Physical organic chemistry, quantified through linear free

energy relationships– Molecules and substituents described by experimentally

observable properties• σ, π, LogP, pKa, LogD, LogS• Abraham’s Parameters

QSAR that lends itself to design by allowing simple interpretation.– The goal is not a better fit, it’s

• A practical working hypothesis• A general explanation


Software Software



Application in today’s world

Explain Activity & Describe Diversity– QSAR, Clustering, Decision trees, Neural Nets, …

Plan Experiments– Library Design (in parallel synthesis or combichem)– Compound selection for screening

Avoid costly errors– Drug-like-ness, Lipinski– ADME prediction– Toxicity alerts– Environmental fate


Software Software



Molecules can be characterized in many different ways

Different forms of input– Whole molecule properties

• pKa• LogP• LogD• Solubility• MW

– Substituent coefficients• Study the impact of substitutions on electronic

structure• Only relevant and effective among series of congeners

– Topological, fingerprints, Bcuts, Keys, etc.• Efficient means to represent chemical structures and

chemistry space within a computer


Software Software



Whole Molecule Properties

ACD has been active in prediction of whole molecule properties since inception nearly 8 years ago– pKa, LogP, LogD, Solubility, BP, VP

Recent commercial highlights– CAS has computed and published properties for over

12Million compounds using ACD PhysChem software– Pfizer has signed a world-wide license for ACD/PhysChem

SoftwareRecent scientific highlights– ACD/LogP now includes the BioByte star list, and has

dramatically increased accuracy– ACD/pKa now includes pKa from BioByte Masterfile,

dramatic improvement in accuracy expected in one year– ACD/Solubility includes AquaSol Database, dramatic

improvements in accuracy coming regularly


Software Software



Substituent Coefficients

Trained from data on 850 substituents with over 3000 carefully derived experimental electronic constants. σ I - 592, σ * (Taft) - 265, σ R - 453, σ R

- - 157, σ R

+ - 143, σ Para - 585, σ Meta - 431,

These same coefficients are used in ACD/pKaPredicted by a fragmental algorithm, similar to ACD/LogP

σ Para- - 142,

σ Para+ - 135,

σ Phosph (P-Acids) -68, σ Ortho (Benzoic Acid) - 41, σ Ortho (Phenol) - 37, σ Ortho (Aniline) - 30, and σ Ortho (Pyridine) - 48


Software Software



Some illustrations

Diversity/Library Design– Benzoic acid library

QSAR– Recent articles from J Med Chem– PLS models done in collaboration with Umetrics

• Sophisticated analysis using only statistically significant components that offer incremental improvement of Q2

– Multilinear regressions done in Excel• Naïve analyses done with manual selection of

variables• No cross validation


Software Software



O

OH

An example: Benzoic acid Library

A simple library based on the benzoic acid core

Draw in the core structure, save it in an SD File


Software Software



ChemFolder view of the Library


Software Software



ACD/Sigma Batch


Software Software



ChemFolder view of molecules and substituent constants


Software Software



ID’s and Sigma’s in Excel


Software Software



Variability in Sigma (Para) substituent Coefficient

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49


Software Software



Results

Notice that Sigma varies significantly with different substituents

Next steps:– Correlation Analysis– Statistical Design of Experiments– QSAR

Output (SDFile or CSV) is compatible with molecular modeling and statistics software packages.


Software Software



Example: QSAR on Antibacterial Activity of Nitrofuran Derivatives

Pires et al. J. Med. Chem 2001, 44, 3673-3681

O

O

H

O

N+

O-

O

R


Software Software




Software Software



PLS: Two significant principal components

0.927 0.821


Software Software



-0.30

-0.20

-0.10

0.00

0.10

0.20

-0.20 -0.10 0.00 0.10 0.20

w*c

[2]

w*c[1]

ACD_LogD_1ACD_LogD_2ACD_LogP

Density

FRB

FW

Index_of_R

MR_MoleculMR_SubstitMV_Molecul

MV_Substit

PSA

Parachor

Polarizabi

Sigma_Ind

Sigma_PHOS

Sigma_Res

Sigma_Res_

Sigma_Res_

Sigma_Star

Sigma_metaSigma_meta

Sigma_orth

Sigma_orth

Sigma_orth

Sigma_orth

Sigma_para

Sigma_paraSigma_para

Surface_Te

Observed

PLS: Loading plot


Software Software



PLS: Loadings in the second component

-0.180

-0.160

-0.140

-0.120

-0.100

-0.080

-0.060

-0.040

-0.020

0.000

0.020

0.040

0.060

0.080

AC

D_L

ogD

_1

AC

D_L

ogD

_2

AC

D_L

ogP

Den

sity

FRB

FW

Inde

x_of

_R

MR

_Mol

ecul

MR

_Sub

stit

MV

_Mol

ecul

MV

_Sub

stit

PS

A

Par

acho

r

Pol

ariz

abi

Sig

ma_

Ind

Sig

ma_

PH

OS

Sig

ma_

Res

Sig

ma_

Res

_

Sig

ma_

Res

_

Sig

ma_

Sta

r

Sig

ma_

met

a

Sig

ma_

met

a

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

para

Sig

ma_

para

Sig

ma_

para

Sur

face

_Te

Coe

ffCS

[2](O

bser

ved)

Var ID (Primary)


Software Software



PLS: Predicted vs Observed

2.30

2.40

2.50

2.60

2.70

2.80

2.90

3.00

3.10

3.20

3.30

3.40

3.50

3.60

3.70

3.80

3.90

4.00

2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00

YVar

(Obs

erve

d)

YPred[2](Observed)

1

2

3

45

67

8

9

y=0.9987*x-0.01038R2=0.9276


Software Software



QSAR for Antibacterial activity of Nitrofuran Derivatives

2

3

4

2 3 4Observed pIC50

Pred

icte

d pI

C50

Multilinearregression

N=9R2=0.91SE=0.19

pIC50=3.96-0.187 LogP

-0.129 σind

-1.289 σres


Software Software



Example: QSAR of OxindoleCDK2 Inhibitors

Neal et al. J. Med. Chem 2001, 44, 4339-4358

NH

N NH

SO

O

NH2

O

R


Software Software




Software Software



PLS: one significant principal component

0.527 0.407


Software Software



-0.40

-0.30

-0.20

-0.10

0.00

0.10

0.20

0.30

0.40

0.50A

CD

_Log

D_1

AC

D_L

ogD

_2

AC

D_L

ogP

AC

D_p

Ka_

Ap

AC

D_p

Ka_

Ap

Den

sity

FRB

FW

Inde

x_of

_R

MR

_Mol

ecul

MV

_Mol

ecul

PS

A

Par

acho

r

Pol

ariz

abi

Sig

ma_

Ind

Sig

ma_

Res

Sur

face

_Te

MR

_Sub

stit

MV

_Sub

stit

Sig

ma_

Res

_

Sig

ma_

Res

_

Sig

ma_

Sta

r

Sig

ma_

met

a

Sig

ma_

met

a

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

orth

Sig

ma_

para

Sig

ma_

para

Sig

ma_

para

Sub

stP

os

Sig

ma_

PH

OS

Act

ivity

1

Act

ivity

2

w*c

[1]

Var ID (Primary)

Loading plot


Software Software



PLS: Predicted vs Observed

0

1

2

3

0 1 2 3

YVar

(Act

ivity

2)(tr

ans)

YPred[1](Activity2)(trans)

21

22

23

24

2526

27

28

29

30

3132

33

34

35

y=1*x+0.00425R2=0.6263


Software Software



QSAR for Oxindole CDK2 Inhibitors

5

6

7

8

9

5 6 7 8 9

Observed pIC50

Pred

icte

d pI

C50

Multilinearregression

N=15R2=0.82SE=0.51

pIC50=5.38+0.346 LogP

-2.561 σind

-1.254 σres

-0.425 #FRB+0.00962MW


Software Software



Conclusions

Now you can consider variation in substituent constants when– Evaluating SAR– Designing Libraries

ACD/Sigma Batch is available– In v6.0 of ACD software– For Win32, Solaris and SGI– As another module for ACD/PhysChem Batch

ACD/Sigma Pro is available for the desktop– One compound at a time.


Software Software



Acknowledgements

ACD– Eduard Kolovanov– Daria Jouravleva

Umetrics– Svante Wold– Mark Earll

Automated Estimation of Hammett Constants - … · • Study the impact of substitutions on electronic structure • Only relevant and effective among series of congeners – Topological,

Documents