Drug Discovery Process Drug Discovery Toolbox Insights on the - - PowerPoint PPT Presentation

SLIDE 1

Insights on the Origins of Biological Activities via Computational Modeling

Associate Professor Dr. Chanin Nantasenamat

Center of Data Mining and Biomedical Informatics! Faculty of Medical Technology, Mahidol University

MU Research Expo 2013 (January 28, 2014) E-mail: chanin.nan@mahidol.ac.th

1

Drug Discovery Process

Ashburn and Thor. Nature Rev. Drug Discov. 3 (2004) 673-683

2

Drug Discovery Toolbox

Combina( torial, Chemistry, Chemical, Libraries, Chemical, Space, High( Throughput, Screening, Property, Filters, Compu( ta;onal, Chemistry, Machine( Learning( QSAR( Proteo3 chemo3 metrics( Molecular( Modeling( Molecular( Dynamics( Molecular( Docking(

3

Experimental activity (pIC50) 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Predicted activity (pIC50) 5.0 5.5 6.0 6.5 7.0 7.5 8.0

What is QSAR? (1)

• !QSAR/QSPR is the

acronym of Quantitative Structure-Activity/Property Relationship!

• QSAR seeks to correlate

structural features of compounds with their biological activities

4

What is QSAR? (2)

• Structure governs activity/

property!

• Typically in the medicinal

chemistry literature, effects

• f substituent groups on

activity is extensively studied

1" 2" 3" 4" 5" 6"

• QSAR/QSPR studies exploits this knowledge for modeling the

biological or chemical activities/properties

5

What is QSAR? (3)

• QSAR involves two main concepts:!
• 1. Generating the physicochemical description!
• 2. Predicting the biological activity or chemical property

Qm# Energy# μ# HOMO# LUMO# HOMO0LUMO#gap# 0.2271& '309.834& 1.0521& '0.21346& '0.0127& 0.20076& 0.2142& '195.31& 0.2337& '0.22611& '0.01915& 0.20696& IC50% 0.05$ 1.50$

Molecular Descriptors Biological Activity Computational Chemistry Machine! Learning

6

Growth of QSAR?

• A search in

SCOPUS shows the growing trend

• f QSAR

publications

7

History of QSAR (1)

1863 Cros Narcotic effects of alcohol increase as solubility of water decrease 1865 Kekulé Chemical structure of benzene 1868 Brown and Fraser Correlation between chemical composition and physiological action 1868 Fischer Lock-and-key concept of enzymes

8

History of QSAR (2)

1893 Richet Discovered that hypnotic effects of organic compounds is inversely related to water solubility 1893 Meyer and Overton Extended the work of Richet Narcotic effects increases with increasing lipophilicity 1937 Hammet

• Proposed the Hammett Equation
• Deduced that ortho, meta, para position of

benzoic acids have effects on the pKa 1958 Kendrew and Perutz Crystal structure of Myoglobin

9

SLIDE 2

History of QSAR (3)

1964 Hansch and Fujita

• Hansch Analysis
• Several physicochemical properties are

combined in one equation

• Extended the earlier work of Hansch and

Muir on plant growth regulators 1964 Free and Wilson

• The concept is easy and no physico-

chemical properties were needed

• Postulated that the biological activity is a

function of its substituent at different positions in a molecule 1977 Karplus Molecular Dynamics of Bovine Pancreatic Trypsin Inhibitor protein 1985 Wuthrich NMR structure of BPT1 protein

10

History of QSAR (4)

1988 Cramer Introduced Comparative Molecular Field Analysis (CoMFA) or what is to be known as 3D-QSAR 1997 Lipinski

• Introduced the Rule of 5 for estimating drug

absorption as well as assess drug-likeness

• Insufficient uptake if MW > 500 Da, LogP > 5,

H-bond donor > 5, H-bond acceptor > 10 1999 Oprea

• Introduced Oprea’s criteria for lead-likeness

2001 Wikberg - Developed Proteochemometrics (so- called Multi-Target QSAR)

• Can be applied at the Proteome level

2003 Congreve - Introduced Rule of 3 for fragment-likeness

11

Applications of QSAR/QSPR models

• !

Regulatory Use: QSAR for modelling environmental toxicity/chemical hazards by EPA and OECD!

• !

Drug Design: QSAR for modelling biological activities!

• !

Materials Design: QSPR for modelling chemical properties

12

GFP$ LPS$ QSAR$ DNA$ PCP$ BPA$

Bacitracin$

Quorum$ Furin$

Vasorelaxant$ Vitamin$E$ Template?$ Monomer$

Phenol$

Sulfonamide$

EDTA?$ DPPC$

Copper$ Complex$ AnDmalarial$ AnD?H1N1$ Aromatase$ Inhibitors$

CYP450$ Inhibitors$

Monte$Carlo$ Feature$$ SelecDon$ Text$ Mining$

13

Big Picture of the QSAR Process

Connec&vity+ Geometrical+ Topological+ Charge+ MLR+

Molecular Descriptors!

• 5.2E+02

9.0E+01 1.3E+03 7.5E+02

• 3.2E+00
• 5.6E-04

9.2E-05

• 4.2E-03

3.1E+00 1.2E+01 9.1E+00 7.8E+00

• 3.4E-01
• 3.0E-04
• 6.5E-11

Molecular Structure! Data Pre-processing! Model Generation! Prediction!

PLS+ PCA+ SOM+ SVM+ NN+ GETAWAY+ WHIM+ 2D+autocorrela&on+ Quantum+Chemical+ Normaliza&on+ Standardiza&on+ KNN+ DT+

Nantasenamat et al. EXCLI J. (2009) 8: 74-88

14

Procedures of QSAR Modeling

Biological(Ac*vity(

• f(Compound(

Series( Chemical( Structures( Geometry( Op*miza*on( Molecular( Descriptors( Machine(Learning( Algorithm( Sta*s*cal( Assessment( Experimental( Valida*on(

15

Biological Activity

• Biological activity of compounds of interest are available from:!
• Primary literature !
• Curated Databases!
• BindingDB!
• MOAD!
• PubChem
• Open Innovation!
• Pharmaceutical companies

are making data publicly available for non- commercial diseases

16

Requirements for compounds

• Compounds belong to a congeneric series!
• Compounds exert the same mechanism of action!
• Compounds bind in a comparable manner

17

Biological activity/chemical property modeled by QSAR

Biological Activity Chemical Property Bioconcentration Boiling point Biodegradation Chromatographic retention time Carcinogenicity Dielectric constant Drug metabolism Diffusion coefficient Inhibitor constant Dissociation constant Mutagenicity Melting point Permeability! Reactivity Blood brain barrier! Solubility Skin Stability! Pharmacokinetics Thermodynamic properties Receptor binding Viscosity

Nantasenamat et al. EXCLI J. (2009) 8: 74-88

18

SLIDE 3

Chemical Structures (1)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

19

Chemical Structures (2)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

20

Chemical Structures (3)

6-phenylsulfonylpyridazin-2H-3-one Clc1ccccc1S(=O)(=O)c1ccc(=O)[nH]n1

Cl S O O O H N N

IUPAC SMILES 2D 3D

21

Computational Chemistry (1)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

22

Computational Chemistry (2)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

23

Computational Chemistry (3)

Nantasenamat et al. Conformational Study of Anastrozole. Unpublished Findings Potential Energy Surface Scan

24

Molecular Descriptors (1)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

25

Molecular Descriptors (2)

Nantasenamat et al. Expert Opin. Drug Discov. (2010) 5: 633-654

26

Molecular Descriptors (3)

27

SLIDE 4

Machine Learning Algorithms (1)

• Machine learning algorithms are used to extract useful

knowledge from data!

• Types of analysis includes regression, classification and

clustering (depending on the objective of the study and nature of the data)

28

Machine Learning Algorithms (2)

• Multiple Linear Regression!
• Partial Least Squares Regression!
• Artificial Neural Network!
• Support Vector Machine!
• Decision Tree

29

Statistical Assessment of QSAR Models

• Common measures on the predictive performance of

QSAR models are as follows:!

• Pearson’s correlation coefficient!
• Root Mean Squared Error!
• F ratio!
• Matthews Correlation Coefficient

30 Experimental Validation of QSAR Models

• Now that we have a predictive QSAR model can validate it

as follows:!

• Selecting novel compounds by rationally designing it or

by selecting one from a large chemical library (e.g. ZINC, PubChem, NCI diversity set, etc.)!

• However, make sure that these compounds fall in the

Applicability Domain of the QSAR model that is the structures are similar to those in the QSAR model

31

Challenges in QSAR modeling

• Although QSAR affords great utility, certain precaution

is necessary as to avoid potential flaws!

• Some inherent drawbacks that undermines the

demise of QSAR have been reported in the literature

32

The Demise of QSAR: Is it reaching a deadend? (1)

• Golbraikh and Tropsha (2002) Beware of q2! J Mol Graph Model 20(4):

269-76.!

• Maggiora (2006) On Outliers and Activity Cliffs: Why QSAR Often
• Disappoints. J Chem Inf Model 46(4):1535.!
• Doweyko (2008) QSAR: dead or alive? J Comput Aided Mol Des 22(2):81-9.!
• Doweyko (2008) Is QSAR relevant to drug discovery? IDrugs 11(12):

894-9.!

• Johnson (2008) The Trouble with QSAR (or How I Learned To Stop

Worrying and Embrace Fallacy). J Chem Inf Model 48(1):25-6.

Problems

33

Dearden’s Top 21 QSAR errors

Dearden et al. SAR QSAR Environ Res 20(3-4):241-66.

34

The Demise of QSAR: Is it reaching a deadend? (2)

• Tropsha et al. (2003) The Importance of Being Earnest: Validation is the Absolute Essential

for Successful Application and Interpretation of QSPR Models. QSAR Comb Sci 22(1):69-77.!

• Polanski et al. (2006) Modeling robust QSAR. J Chem Inf Model 46(6):2310-8.!
• Tropsha and Golbraikh (2007) Predictive QSAR modeling workflow, model applicability

domains, and virtual screening. Curr Pharm Des 13(34):3494-504.!

• Dearden et al. (2009) How not to develop a quantitative structure-activity or structure-

property relationship (QSAR/QSPR). SAR QSAR Environ Res 20(3-4):241-66.!

• Nantasenamat et al. (2009) A practical overview of quantitative structure-activity
• relationship. EXCLI J 8:74-88.!
• Nantasenamat et al. (2010) Advances in computational methods to predict the biological

activity of compounds. Expert Opin Drug Discov 5(7)633-54.

Solutions

35

The Demise of QSAR: Is it reaching a deadend? (3)

• Tropsha (2010) Best Practices for QSAR Model Development, Validation,

and Exploitation. Mol Inf 29(6-7):476-88.!

• Ajain (2010) QMOD: physically meaningful QSAR. J Comput Aided Mol Des

24(10):865-78.!

• Scior (2012) Recognizing Pitfalls in Virtual Screening: A Critical Review. J

Chem Inf Model 52(4):867-81.!

• Golbraikh (2014) Data Set Modelability by QSAR. J Chem Inf Model, DOI:

10.1021/ci400572x.!

• Cherkasov et al. (2014) QSAR Modeling: Where Have You Been? Where Are

You Going To? J Med Chem DOI: 10.1021/jm4004285.

Solutions

36

SLIDE 5

Major Limitations of QSAR/QSPR

• Although immensely useful, QSAR/QSPR encompasses the study of a

series of ligands against a single protein target!

• As such, if a series of ligands were to be investigated against a family of

related targets (so-called multi-target) then this would involve the construction of several QSAR/QSPR models against each of the targets!

• The concept of multi-target is highly relevant for phenomenons as

polypharmacology, off-target binding, drug side-effects, drug promiscuity,

• rphan receptors for which no drugs exists, etc.!
• These problems can be solved by Proteochemometrics

37

Choice of Molecular Descriptor

• !It should be noted that molecules in a library are not

described purely by their structural features but rather by their ability to interact with target proteins of interest!

• !The concept of structure-activity cliff had been

proposed by Bajorath to explain why structurally similar molecules may have drastic differences in activity

38 Interpolation/Extrapolation of QSAR Model

• It is quite tempting to apply the learned quantitative

relationships to predict new chemical structures having better properties than those used to formulate the original QSAR!

• However, as QSAR is essentially a statistical analysis, therefore

it should be reminded that the modeled results only serve to characterize the trends of properties within the confinement

• f the applicability domain (that is the constituting data

samples used to train the model)

39

QSAR Model Reliability

• Given that a QSAR model has high accuracy, the

question is how can we be sure of its reliability!

• Statistical parameters are commonly used to assess

their statistical robustness such as Pearson’s correlation coefficient, Root mean squared error, etc.

40 Model Accuracy versus Model Interpretation

• !

Sometimes models with high accuracy may employ

• bscure or difficult to understand molecular

descriptors while models with lower accuracy may use easier to understand molecular descriptors

41

Complementary Technologies augmenting QSAR

• Combinatorial Chemistry enables the synthesis of large volume of

chemicals!

• High-Throughput Screening platforms for assaying biological activities/

chemical properties on a large-scale!

• Novel Machine Learning Algorithms!
• Automation in Workflow via programming/non-programming

approaches!

• Uniformity/Standardization of methodology

42

Proteochemometrics

• Proteochemometrics was developed by Maris Lapins and Jarl Wikberg of

Uppsala University in 2001!

• Advantages!
• Can explain ligand-target affinity by providing detailed maps down to

the substructures and amino acid level!

• Can be used to rationalise why a ligand is active toward one target and

not on the other related target!

• Has been shown to be useful for Drug Repositioning!
• Could be used for Personalized Medicine

43

Conclusion (1)

• It is without a doubt that the QSAR paradigm boasts much benefit for the rational design
• f robust compounds!
• Nevertheless, there are certain shortcomings that may limit the widespread application
• f QSAR!
• Workflow of QSAR model development!
• High dimensionality of the input space!
• Representation of the molecular structure!
• Interpretability and meaning of the developed QSAR models!
• Presence of outliers or activity cliffs!
• Validation of QSAR model performance!
• Applicability in real-world setting

44

Conclusion (2)

• In spite of certain inherent flaws, the QSAR paradigms inevitably
• ne of the most useful forces contributing to the rapid

development of drug discovery and design.!

• As with all technologies, QSAR is not perfect; however, its

weaknesses and flaws are continuously being identified, solved and reformed to help shape a new improved and robust approach that is approaching minimal predictive error!

• To help realize the goal of developing an intuitive approach

toward the development of robust QSAR models, our laboratory had developed a software that affords a semi- automated if not automated QSAR modeling.

45

SLIDE 6

Conclusion (3)

• At nearly 10 years of QSAR research, we can say that the

demise of QSAR is a myth if done properly and we had only scratched the surface of its full potential!

• QSAR is continuously evolving…starting from 2D-QSAR to 


8D-QSAR!!

• Proteochemometrics (so to say Multi-Target QSAR) enables

us to take advantage of the explosion of Omics data

46

A"so%ware"for"performing"automated"Data"Mining" AutoWeka"is"a" Python"wrapper"

• f"Weka"
• It is freely available at http://www.mt.mahidol.ac.th/autoweka/

47

Acknowledgements

Mahidol University

• Prof. Dr. Virapong Prachayasittikul!
• Prof. Dr. Supaluk Prachayasittikul!
• Prof. Dr. Chartchalerm Isarankura-Na-Ayudhya!
• Dr. Apilak Worachartcheewan!
• Dr. Wiwat Owasirikul!
• Dr. Watshara Shoombuatong!
• Prasit Mandi!
• Hao Li!
• Naravut Suvannang!
• Sunanta Nabu

Uppsala University

• Prof. Dr. Jarl Wikberg!
• Dr. Maris Lapins

Lund University

• Prof. Dr. Leif Bulow

Financial Support

• Mahidol University!
• National Research Universities Grant!
• Uppsala University!
• Swedish Research Council (VR)

48