QSAR and pharmacophore modelingfch.upol.cz/wp-content/uploads/2015/11/QSAR... · QSAR and...

QSAR and pharmacophore modeling

Pavel Polishchuk, Ph.D.Karel Berka , Ph.D.

Jindřich Fanfrlík , Ph.D.Martin Lepšík , Ph.D.

Institute of Molecular and Translational MedicinePalacky University

[email protected]

Advanced in silico drug design workshopPalacky Univrsity, Olomouc, Czech Republic 30 January – 1 February 2017

Quantitative structure-activity relationship (QSAR) modeling

Activity = F(structure)

M – mapping functionE – encoding function

Modeling of compounds properties

Activity = M(E(structure))

X1 X2 X3 X4 X5 X6 … XN

1 0 9 0 11 1 … 1

4 0 1 0 0 0 … 1

0 0 0 0 0 4 … 6

0 2 3 6 0 0 … 3

… … … … … … … …

4 0 0 0 1 2 … 1

QSAR modeling workflow

X1 X2 X3 X4 X5 X6 … XN

1 0 9 0 11 1 … 1

4 0 1 0 0 0 … 1

0 0 0 0 0 4 … 6

0 2 3 6 0 0 … 3

… … … … … … … …

4 0 0 0 1 2 … 1

StructureDescriptors(features) Model

Encoding(represent structure with

numerical features)

Mapping(machine learning)

Y

1.1

1.4

6.8

3.0

…

1.5

End-pointvalues

Overall QSAR workflow

Input data

Bioassays

Databases

PreprocessingFeature

engineeringModel

learningModel

validation

Classification

Regression

Clustering

Cross-validation

Bootstrap

Test set

Applicability Domain

Feature selection

Feature combination

Data normalization

Feature extraction

Interpretation

j

j

i

i

z

xxx

'

Step 1. Data collection

Scientific literature and patentsDatabases (ChEMBL, PubChem, BindingDB, etc)Own experimental data

Traditionally modeled compounds should have the same mechanism of action, however using of complex non-linear machine learning method allows to model data sets with mixed or even unknown mechanism of action with reasonable accuracy.

Conditions may substantially influence the results of bioassays (change in temperature, activators, detectors, etc)

Units checking

Each found error is next to last

Step 2. Data curation (normalization)

strange units


Data from NCI60

S is charged

overall charge is +2

overall charge is +2HClO4 is represented with separated chargesnitrogens are covalently bond to Zn?wrong stoichiometry?

unusual valence of S


Removal of mixtures, inorganics, metalorganics, etc

Strip of salts, counterions, etc

Ionization, if necessary (at the particular pH level)

Chemotype normalization, resonance structure and tautomers

Duplicates removal

Manual checking

Step 3. Descriptors: classificationObject type:

molecular descriptors (single molecules)descriptors of molecular ensemble (mixtures, materials)reaction descriptors (reactions)

Descriptor origin:calculated from the structureempirical (Hammet constants, chemical shifts in NMR, etc)

Locality:local (atom charge)global (molecular weight, molecular volume, lipophilicity, etc)

Dimensionality:1D (number of methyl groups, molecular weight, etc)2D (topological indices, fragmental descriptors)3D (molecular volume, quantum chemical descriptors)4D (based on a set of conformers)

Calculation method:physico-chemical (lipophilicity, etc)topological (invariants of molecular graph, Randic index, Wiener index, etc)fragmental (fingerprints, etc)pharmacophorespatial (moment of inertia, etc)quantum-chemical (energy of HOMO/LUMO, etc)etc. R. Todeschini and V. Consonni Handbook of Molecular Descriptors, 2008

Atom-centric (augmented atoms) fingerprints

Generate substructures starting from each atom and considering all its neighbors up to the specified distance (radius or diameter).

Morgan fingerprints, Extended-connectivity fingerprints (ECFP). Functional-class fingerprints (FCFP) , etc.

radius=2 (diameter=4)

Rogers, D. & Hahn, M. Extended-Connectivity Fingerprints. J. Chem. Inf. Model. 50, 742-754 (2010)

Morgan fingerprintsradius=2

(diameter=4)

Morgan fingerprintsradius=4

(diameter=8)

identical fingerprints

different fingerprints

Atom-centric (augmented atoms) fingerprints

Fingerprints

Each molecule has variable length set of substructures – variable length fingerprints

N-C-C C-C-O C-C=O C-C-C C:C:C C:C:N C:N:C

1 1 1 0 0 0 0

1 1 1 1 0 0 0

0 0 0 0 1 1 1

2-bond sequences

Counts of each fragment can be used instead of binary fingerprints

Hashed fingerprints

Have fixed length (usually 512, 1024 or 2048 bits)

hash code

N-C-C C-C-O C-C=O C-C-C C:C:C C:C:N

13823 9740 37278 28478 874 283764

0 0 0 1 0 0 1 1 0 1 0 0 1 0 1 0 0 0 1 1

pseudo-randomnumber generator

fixed-length bit string

Each substructure activates several bits (usually 4-5) to avoid collisions and produce bit string of enough density

Missing bits mean that certain substructures are not presented, Active bits mean that certain substructure may be present (but due to possible collisions one cannot be sure)

Step 4. Feature processing

Feature transformations:linear and non-linear scaling

Feature combinations:

Feature selection:removal of correlated descriptors (important e.g. for linear regression models)removal of descriptors with missing valuesremoval constant and near constant featuresremoval of descriptors with low correlation with the target propertystep-wise feature selectionevolutionary feature selection (genetic algorithm, etc)

sd

xxz

i

i

n

i

ixx

n

nsd

1

2)(

1

ixi

ez

1

1range (0; 1)

jiijxxz

2

iixz add quadratic term

Step 5. Model building

Unsupervisedclustering

Supervised

Regression Classification

Multiple linear regression (MLR)

Partial linear regression (PLS) Logistic regression

Gaussian Process (GP) Naïve Bayes (NB)

Decision trees (DT)

Support vector machine (SVM)

Neural nets (NN)

Random forest (RF)

k-Nearest neighbors (kNN)

1/C = 4.08π – 2.14π2 + 2.78σ + 3.38

π = logPX – logPH

σ - Hammet constant

plant growth inhibition activity of phenoxyacetic acids

Hansch equation

R is H or CH3;

X is Br, Cl, NO2 and

Y is NO2, NH2, NHC(=O)CH3

Inhibition activity of compounds against Staphylococcus aureus

Act = 75RH – 112RCH3 + 84XCl – 16XBr – 26XNO2 + 123YNH2 + 18YNHC(=O)CH3 – 218YNO2

Free-Wilson models

QSAR model building

18

Simulated data set of actives and inactives with two descriptors – MW and LogP

LogP

MW

Decision tree

Decision tree

LogP

MW

LogP < 3.2YES NO

LogP

MW

LogP

MW

MW < 358YES NO LogP >= 4.4YES NO

IF LogP >= 3.2 AND LogP < 4.4 a compound is

20

Simulated data set of actives and inactives with two descriptors – MW and LogP

LogP

MW

Decision tree

Initial dataset

Bootstrapsample

Bootstrapsample

Bootstrapsample

CART tree1 CART tree2 CART tree3

Combined prediction

…

Random feature subspace in each node

Random Forest

Consensus (ensemble) modeling

Data set

Random Forest

Support vector

machine

Linear regression

Neural network…

Combined predictions

Models should be not correlated(one may use different combination of descriptors and machine learning methods)

averaging/major voting

Consensus (ensemble) modeling

Data set

Random Forest

Support vector

machine

Linear regression

Neural network…

Metamodel(linear regression)

stacking

Prediction

Step 6. Validation

Test set (usually 20-25% of the work set)

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2working set

random test set 1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2

stratified test set

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2

Cross-validation

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2working set

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2

1.2 1.3 1.7 2.0 2.2 2.8 3.1 3.2 3.2 3.6 4.7 5.7 5.8 6.4 7.2 8.1 9.0 9.1 9.2

fold 1

fold 2

fold 3

predictions of different folds are combined to calculate the final predictive measure

Step 6. Measures of predictive ability of models

Classification

FPTN

TNySpecificit

FNTP

TPySensitivit

2

ySensitivitySpecificitaccuracy Balanced

baseline-1

baseline -accuracy Kappa

N

TNTPaccuracy

2N

FP)FN)(TP(TPFN)FP)(TN(TNbaseline

Confusion matrix

Predicted

positive class (1)

negative class (0)

observed

positive class (1)

truepositive

(TP)

false negative

(FN)

negative class (0)

false positive

(FP)

true negative

(TN)

))()()(( FNTNFPTNFNTPFPTP

FNFPTNTPMCC

[-1; 1]

[0; 1]

[0; 1]

[0; 1]

[0; 1]

[0; 1]

Step 6. Measures of predictive ability of models

Regression

i

2

obspredi,

i

2

obsi,predi,

2

)y(y

)y(y

1Q

1N

)y(y

RMSEi

2

obsi,predi,

N

1i

obsi,predi,yy

N

1MAE

Determination coefficient

Root mean squared error

Mean absolute error

Step 7. Applicability domain (AD)Extrapolation to very distant objects is dangerous

There is a need to define the domain where our model is reliable (models are not universal!)

Only compounds which are similar to the training set compounds should be included in applicability domain of the model. One should estimate similarity of new compounds (test set, etc) to the training set compounds.

Step 7. Applicability domain (AD) measures

Bounding box - based on descriptor range

Desc_1

Desc_2

AD• internal regions are usually empty, especially if the number of descriptors is big

• it doesn’t take into account descriptor correlation

Distance from training set compounds in descriptor space

Desc_1

Desc_2Distance from training set compounds in model space

Model_1

Model_2

Requires several models (e.g. consensus model, bootstrap models)

One-class SVM

Conformal predictions

Step 7. Applicability domain (AD) example

J. Chem. Inf. Model., Vol. 50, No. 12, 2010

Step 8. Interpretation of QSAR models

Found active/inactive patterns can be used for optimization of compound properties

Retrieved trends of structure-activity relationships can be used for knowledge-based validation of QSAR models

Knowledge mining with the analysis of interpretation results – find new structure-activity relationships

Regulatory porposes

Why interpretation is important?


Principles and issues

Model should be predictive

Interpretation is valid within the applicability domain of the model

Interpretation results are data set dependent


1/C = 4.08π – 2.14π2 + 2.78σ + 3.38

π = logPX – logPH

σ - Hammet constant

plant growth inhibition activity of phenoxyacetic acids

electronic factorsrate of penetration of membranes in the plant cell

Hansch equation

R is H or CH3;

X is Br, Cl, NO2 and

Y is NO2, NH2, NHC(=O)CH3

Inhibition activity of compounds against Staphylococcus aureus

Act = 75RH – 112RCH3 + 84XCl – 16XBr – 26XNO2 + 123YNH2 + 18YNHC(=O)CH3 – 218YNO2

Free-Wilson models

-2

-2

-1

0

2

4

4

1


2 δ

δ)f(xδ)f(xC

f(x)δ)f(xC

δ)f(xf(x)C

forward difference

backward difference

central difference

Partial derivatives is used to calculate contributions of single features.It is applicable to any model built on meaningful (interpretable) descriptors.May be calculated analytically or numerically.

Estimated contributions of separate substructural features can be mapped back on the structure to reveal contribution of fragments.

C-C-N -3

C-C-O +6


347 agonists of 5-HT1A receptor

Ar - substituted (hetero)arylsL - polymethylene chainR - various (poly)cyclic residues

O OO M e C l C l C F

3

N

N

C H3

F

O2N

C l

PLS 0.84 0.18 0.03 -0.04 -0.06 -0.09 -0.11 -0.66 -0.73 -0.94 -0.96RF 0.27 0.24 0.04 0.07 -0.02 0.11 0.04 -0.04 -0.55 -0.66 -0.66

Ar

L -(CH2)6- -(CH2)5- -(CH2)4- -(CH2)3- -(CH2)2- -CH2-

PLS 0.8 0.71 0.81 0.08 -0.04 0.06

RF 0.14 0.19 0.14 -0.01 -0.03 0.05

RF built on 6460 fragmental descriptorsPLS built on a subset of 72 of those descriptors


=–

A B C

Activitypred(A) Activitypred(B) Contribution(C)

f(A) = x f(B) = y W(C) = x – y

Polishchuk, P. G.; Kuz'min, V. E.; Artemenko, A. G.; Muratov, E. N. Universal Approach for Structural Interpretation of QSAR/QSPR Models. Molecular Informatics 2013, 32, 843-853

Universal approach


Polishchuk P, Tinkov O, Khristova T, Ognichenko L, Kosinskaya A, Varnek A, Kuz’min V. Journal of Chemical Information and Modeling 2016, 56, 1455-1469.

Acute oral toxicity on rats

?

?


Descriptors Algorithm Balanced Accuracy

SiRMSRF 0.817

SVM 0.800

DragonRF 0.816

SVM 0.793

5-fold external cross validation results

Polishchuk P.G. et al. Molecular Informatics, 2013, 843-853

Chlorine Bromine Iodine

Den

sity

ContributionContributionContribution

Aliphatic halogens (F < Cl < Br < I)

unpublished results


Toxicity towards Tetrahymena pyriformis


Mean

OH (aliphatic) -0.33

OH (aromatic) 0.25

COOH (aliphatic) 0.01

COOH (aromatic) -0.45

Contribution

Den

sity

Context dependence of contribution to toxicity of OH and COOH group

Toxicity towards Tetrahymena pyriformis

Case

number

Do specific interactions of a ligand

with its target exist or important?

Is an orientation of

a ligand relatively

its target known?

Fragments selection

and grouping

1

NO

(e.g. passive diffusion through mem-

branes, solubility, lipophilicity, etc)

not relevant

can be done by the

researcher based on his

own knowledge

2

YES

(ligand-receptor interactions,

host-guest complexes, etc)

YES

consider fragments’

positions relatively to

the target and observed

or predicted

interactions

3 NO

MMP can be applied,

silently assumed that all

compounds have the

same interaction mode


Step 8. Interpretation of QSAR models: conclusions

Interpretation results of a good predictive model should converge independent of:

interpretation approachdescriptorsmachine learning method

All models are interpretable but not all end-points

Overall QSAR workflow

Input data

Bioassays

Databases

PreprocessingFeature

engineeringModel

learningModel

validation

Classification

Regression

Clustering

Cross-validation

Bootstrap

Test set

Applicability Domain

Feature selection

Feature combination

Data normalization

Feature extraction

Interpretation

j

j

i

i

z

xxx

'

Pharmacophore modeling

2.4 Å

estradiol and trans-diethylsilbestrol.

14.5 Å

2.3 Å

6.7 Å 6.9 Å

44

Early pharmacophore models

ionic interaction

H-bond

π‒π interaction

ionic interaction

H-bond

π‒π interaction

X

A B

45

(R)-(-) –Epinephrine(Adrenalin)

(S)-(+) –Epinephrine

Early pharmacophore models

A pharmacophore is the ensemble of steric and electronic features that is necessary to ensure the optimal supramolecular interaction with a specific biological target structure and to trigger (or block) its biological response.

Annu. Rep. Med. Chem. 1998, 33, 385–395

46

Pharmacophore definition

UniversalPharmacophore models represent chemical functions, valid not only for the currently bound, but also unknown molecules

Computationally EfficientDue to their simplicity, they are suitable for large scale virtual screening (>109 compounds, also in parallel settings)

Comprehensive & EditableSelectivity-tuning by adding or omitting chemical feature constraints, information can be easily traced back

47

Advantages of pharmacophore models

Atom- and pharmacophore-based alignment

Methotrexate Dihydrofolate

Hydrogen bonding patterns

Atom-based alignment Pharmacophore alignment

Features: Electrostatic interactions, H-bonding, aromatic interactions, hydrophobic regions, coordination to metal ions ...

49

Feature-based pharmacophore models

H-bond donor

H-bond acceptor

Positive ionizable

Negative ionizable

Hydrophobic

Aromatic ring

Pharmacophore features

Ligand-based pharmacophore

Exploration of conformational space

Multiple superpositioning experiments

DISCO, Catalyst, Phase, MOE, Galahad, LigandScout ...

Structure-based pharmacophore

GRID interaction fields: Convert regions of high interaction energy into pharmacophore point locations & constraints[S. Alcaro et al., Bioinformatics 22, 1456-1463, 2006]

Start from target-ligand complex: Convert interaction pattern into pharmacophore point locations & constraints[G. Wolber et al., J. Chem. Inf. Model. 45, 160-169, 2005]

50

Ligand- and structure-based pharmacophores

Typical pharmacophore modeling workflow

Clean structures of ligands

Generate conformers

Assign pharmacophore features

Find common pharmacophores

Score pharmacophore hypotheses

Validation of pharmacophore models

LS: Comparability vs Specificity of Chemical FeaturesLevels Universality Specificity Classification Example

1 -- +++

Molecular graph descriptor (atom, bond) with geometric constraint

A phenol group facing a parallel benzenoidsystem within a distanceof 2–4A

2 - ++

Molecular graph descriptor (atom, bond) without geometric constraint

A phenol group

3 ++ +

Chemical functionality (hydrogen bond donor,acceptor) with geometric constraint

H-bond acceptor vector including an acceptor point as well as a projected donor point; aromatic ring including a ring plane

4 +++ -

Chemical functionality (positive ionizable area, lipophilic contact) without geometric constraint

H-bond acceptor without the projected point; lipophilic group

LigandScout SMARTS pharmacophore patterns

Common features finding

H-bond donor

H-bond acceptor

Negative ionizable

Hydrophobic

Pharmacophore features

55RET Kinase Inhibitors

2IVU

2IVV

Shared consensus pharmacophore (LigandScout)

56RET Kinase Inhibitors

2IVU

2IVV

Merged consensus pharmacophores (LigandScout)

Validation with external test set

57

Observed

0 1

Pre

dic

ted 0 TN FN

1 FP TP

Pharmacophore validation

Negatives Positives

1

InactivesActives

Nu

mb

er o

f co

mp

ou

nd

s

Threshold Si Fit score 0 11-Sp

Se

AUC = 0-1

InactivesActives

Nu

mb

er o

f co

mp

ou

nd

s


Se

AUC = 1

Inactives

Actives

Nu

mb

er o

f co

mp

ou

nd

s


Se

1

AUC =0.5

58

1

Pharmacophore validation: ROC

Decision making from ROC curves

1

0 11-Sp

Se

InactivesActives

Break-even strategyOptimize both Se and Sp

S2

Inactives

Actives

Conservative strategyIncrease the hit rate

Faster, cheaper, motivating

S3

InactivesActives

Liberal strategyIncrease chemical diversity

Innovative

S1

59

Pharmacophore validation: ROC

60

Liga

nd

s

Pharmacophore models

Pharmacophore: ligand profiling

T. Steindl et al., J. Chem. Inf. Model., 46, 2146-2157 (2006)

61

8.05 Å7.36 Å

5.6 Å

Fragment-based library designAntagonists of thromboxane A2 receptor

Khristova T., et al., Reports of NAS of Ukraine, 2014, 103-108

62

Choosing of the training set compounds (ligand-based) or complexes (structure-based)

Ligand-based:generation of conformers, choosing of “bioactive” conformer and reference molecule

Structure-based:different interaction patterns of different ligands;high specificity of models

Common problems of pharmacophore

63

Pharmacophores and Pharmacophore SearchesEds.: Thierry Langer, Rémy D. Hoffmann2006

Recommended literature

64

J. Chem. Inf. Model., 2012, 52 (6), pp 1607–1620

Recommended literature

QSAR and pharmacophore modelingfch.upol.cz/wp-content/uploads/2015/11/QSAR... · QSAR and...

Documents

Transcript of QSAR and pharmacophore modelingfch.upol.cz/wp-content/uploads/2015/11/QSAR... · QSAR and...