LigEvolver: A Tool for ligand formulation using genetic algorithm

Rochester Institute of Technology

RIT Scholar Works

Theses

Thesis/Dissertation Collections

1-2006

LigEvolver: A Tool for ligand formulation using

genetic algorithm

Priyadarsini Soundararajan

Follow this and additional works at:

http://scholarworks.rit.edu/theses

This Thesis is brought to you for free and open access by the Thesis/Dissertation Collections at RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please [email protected].

Recommended Citation

LigEvolver: A Tool for Ligand Formulation Using Genetic

Algorithm

Approved: _ _

P_a_u_I_A_._C_r

a---=i

9=----__

G. R. Skuse

Head, Department of Biological Sciences

Submitted in partial fulfillment of the requirements for the Master of Science

degree in Bioinformatics at the Rochester Institute of Technology.

Thesis/Dissertation Author Permission Statement

Ti tIe of thesis or dissertation:

_--"

l===luf:.zJ.'2Z~e~\J~O~L<JlV--bf:Rd::>..~

~

---I

A

~---,

'

fc!..10~O:!..!L

""--_

EI--..\.<Oh.R

..----J

L""'-..lI.I.J('!I[JJA~NuD

L---Name of author: &1L(ADBRsI N\

~\

N"smR..

Of

Program: l?lbII\lFDR.\-j t\IICS

College:

Sc/f

=-

NCfe

I understand that I must submit a print copy of my thesis or dissertation to the RIT Archives, per current RIT guidelines for the completion of my degree. I hereby grant to the Rochester Institute of Technology and its agents the non-exclusive license to archive and make accessible my thesis or dissertation in whole or in part in all forms of media in perpetuity. I retain all other ownership rights to the copyright of the thesis or dissertation. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

Print Reproduction Permission Granted:

I, fBNfiDAfS5lNI

c

~

O

N'D8f<tlRf\rAN

Signature of Author: _ _ _

P_r_iy~a_d_a_r_s_i_n_i

___

1

i

/&4/0

b

Print Reproduction Permission Denied:

I, , hereby deny permission to the RIT Library of the Rochester Institute of Technology to reproduce my print thesis or dissertation in whole or in part.

Signature of Author: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Date: _ _ _ _ _

Inclusion in the

Digital Media Library Electronic Thesis

Dissertation (ETD) Archive

I, ' additionally grant to the Rochester Institute of Technology Digital Media Library (RIT DML) the non-exclusive license to archive and provide electronic access to my thesis or dissertation in whole or in part in all forms of media in perpetuity.

I understand that my work, in addition to its bibliographic record and abstract, will be available to the world-wide community of scholars and researchers through the RIT DML. I retain all other ownership rights to the copyright of the thesis or dissertation. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I am aware that the Rochester Institute of Technology does not require registration of copyright for ETDs.

Thesis

Advisory

Committee Chair

Dr. Paul_Craig

Professorof_Chemistry

Departmentof_Chemistry

CollegeofScience

Rochester Instituteof_Technology

Committee Member

Professor Al Biles

ProfessorandUndergraduate Program Coordinator

Information_TechnologyDepartment

Golisano Collegeof_ComputingandInformation Sciences

Rochester Instituteof_Technology

Committee Member

Dr. Anne Haake

Associate Professor

Information_TechnologyDepartment

Golisano Collegeof_ComputingandInformationSciences

Abstract

Human

Immunodeficiency

-1)has been foundtobe thecause of

Acquired Immune _Deficiency Syndrome (AIDS). This virus caused uproar _during the

1980's and started a major _{drug discovery} race _among the pharmaceutical companies. It

also proved to be a milestonein theestablishmentoftheuse of computers inrational _drug

design.

Evolutionary computation is a _commonly used computational approach that has

been successfully applied to a _variety of fields ranging from _engineering to life science.

The main reason for its effectiveness is that it is driven _by the principles of evolution. A

genetic algorithm is an approach to _evolutionary computation that allows random

combination of data to occur in a series of generations and enables the identification of

novel systemsthatmight otherwisehavegone undetected.

This work explores the use of genetic algorithms to generate new ligand structures

that_{may be} effectivein _inhibitingHIV- 1

Protease, one ofthemajor_drug targets in HIV.

One of the computational challenges associated with _{drug discovery} is the conversion of

chemical and biological entities into formats that the computer can use. Chemical

structures can be represented _by linear character strings called SMILES strings. SMILES

(Simplified Molecular Input Line Entry Specification) strings are taken as the genetic

representation for our approach to _drug discovery, and a genetic algorithm has been

the ligands are evaluated and either kept or removed from the gene pool _following the

Acknowledgements

I would like to thank Dr. Paul _Craig, committee chairman and advisor for his_unrelenting

guidance and encouragement. His patience and enthusiastic supervision has been the

driving force behind the successful completion of the project. I am grateful to Prof. Al

Biles for his zeal andsupport. Thenumerous technicaldiscussions withhimenabled me to

approach the problem in various angles and his advices towards the computational aspect

oftheresearch were useful. I would like to offer _my sincere gratitude to Dr. Anne Haake

for herassistance andinterest.Dr. Haake had been a source of supportand encouragement

at all times. Iwould like to thankallthe committee members for_trusting me and _being a

part ofthis thesis.

I would like to extend _my thankfulness to Mr. Srinivas _{Naidu, my} uncle and a constant

Table ofContents

1. Introduction 1

1.1 Fundamental Concepts 1

1.2 PurposeofInvestigation 2

1.3Problem Description 3

1.4Rationale 4

2. MethodsandMaterials 5

2.1 DiseaseandEnzyme Selection 5

2.1.1 HIVGenome 5

2.1.2Lifecycle ofHIV 6

2.1.3 HIVProtease 8

2.1.4ConservedRegion 9

2.1.5 EnzymaticReaction 10

2.2 Genetic RepresentationinLigEvolver 12

2.2.1 Chromosome Representation 12

2.2.2Sample SMILES _string and2D structure ofligand 13

2.2.3 PharmacophoricFeaturesConsidered 14

2.3Input 15

2.3.1 XML/Perl Parser 16

2.3.2 Protein Data Bank 17

2.4 Output 17

2.6.1 Genetic Algorithm 19

2.6.2 Pharmacophore Elucidator 27

2.6.3 Molecular Weight Calculator 32

2.6.4 Van Der Waals Volume 33

2.6.5 Polar Surface Area 35

2.6.6 Motif Finder 35

2.6.7 Fitness Computor 36

3. Results 41

3.1 Simulation 1 41

3.1.1 Significant Results 41

3.2 SimulationII 46

3.2.1 Significant Results 46

3.3 Simulation HI 52

3.3.1 Significant Results 52

3.4 Simulation IV 58

3.4.1 Significant Results 58

3.5Simulation V 61

3.5.1 Significant Results 61

3.6SimulationVI 64

3.6.1 Significant Results 65

3.7 FitnessBehavior 66

5. Conclusion 72

Bibliography 73

Appendix 1 78

ListofFigures

Figure 2.1.1.1 HIV Genome 5

Figure 2.1.2.1 HIV Lifecycle 7

Figure 2. 1.3.1 HIV Protease and active_site 8

Figure 2.1.3.2HIV Protease inhibited

by 1D4Y 8

Figure 2.1.4.1 Asparatic Acid 9

Figure 2.1.4.2Threonine 9

Figure 2.1.4.3 Glycine 9

Figure 2. 1.4.4Catalytictriad 1 0

Figure2.1.5.1 HIV Protease_cleavingreaction 11

Figure 2.2.1.1 Transformationfrom3D structuretoSMILES _string 12

Figure2.2.2.2TPVStructure 14

Figure2.2.3.1 TPV Structurewithlabeledpharmacophoricfeatures 15

Figure2.3.1 Input Construction 17

Figure 2.6.1 ComponentsofLigEvolver 19

Figure 2.6.1.1 Flowchart ofGA 20

Figure 2.6.1.1.1.1 Exampleof_GenFrag 22

Figure 2.6.1.1.2.1 ExampleofCrossover 23

Figure2.6.1.1.3 Mutationrates and genetic_adaptability 24

Figure 2.6.1.2.1 Tournament Selection 26

Figure 2.6.2.1.1 Aromatic StructureI

- ABenzene

ring 27

Figure2.6.2.2. 1 Hydrogen Bond AcceptorStructure I 28

Figure 2.6.2.2.3 HydrogenBond Acceptor StructureIII 28

Figure 2.6.2.2.4 Hydrogen Bond Acceptor StructureIV 29

Figure 2.6.2.3. 1 Hydrogen Bond A/D Structure I 29

Figure 2.6.2.3.2 Hydrogen Bond A/D Structure II 29

Figure 2.6.2.3.3 Hydrogen BondA/D StructureIII 30

Figure 2.6.2.4. 1 HydrophobicStructure I 30

Figure 2.6.2.4.2 Hydrophobic StructureII 30

Figure 2.6.2.4.3 Hydrophobic StructureIII 3 1

Figure 2.6.2.4.4 Hydrophobic StructureIV 3 1

Figure 2.6.2.4.5 Hydrophobic Structure V 3 1

Figure 2.6.2.4.6 Hydrophobic Structure VI 32

Figure 2.6.2.4.7 Hydrophobic StructureVII 32

Figure3.1.1.1.1.1 Simulation I- Generation 100- CompoundI

42

Figure 3.1.1.1.1.2 SimulationI- Generation 100- Compound II

43

Figure 3.1.1.2.1.1 SimulationI- Generation 85

-CompoundI 44

Figure3.1.1.2.2.1 Simulation I

- Generation 85- CompoundII

45

Figure 3.2. 1.1.1.1 Simulation II

- Generation40

-CompoundI 47

Figure 3.2.1.1.2.1 Simulation II

- Generation

40- Compound II

48

Figure3.2.1.1.3.1 SimulationII

- Generation 40- Compound HI

49

Figure 3.2. 1.1.4.1 SimulationII

- Generation 40- Compound IV

50

Figure3.2. 1.2.1.1 SimulationII

- Generation41 - Compound I

5 1

Figure3.3.1.1.1.1 SimulationHI- Generation 100- Compound I

(a) 53

Figure 3.3.1.1.1.2SimulationIII- Generation 100- Compound I

Figure 3.3.1.1.1.3 Simulation III

-Generation 100 - CompoundI

(c) 55

Figure 3.3.1.1.2.1 SimulationHI

-Generation 100 - CompoundII

(a) 56

Figure 3.3.1.1.2.2 SimulationIII- Generation 100- CompoundII

(b) 57

Figure 3.3.1.1.3.1 Simulation III- Generation 100 - CompoundIII 57

Figure 3.4.1.1.1.1 SimulationTV- Generation 100

-Compound I 59

Figure 3.4.1.1.2.1 SimulationIV

-Generation 100- CompoundII 60

Figure 3.5.1.1.1.1 SimulationV

-Generation 12- Compound I 62

Figure 3.5.1.2.1.1 SimulationV

- Generation 10- CompoundI 63

Figure 3.5.1.3.1.1 SimulationV- Generation 9- Compound

I 64

Figure 3.6.1.1.1.1 SimulationVI

-Generation 199- CompoundI 65

Figure 3.7.1 Fitness Behavior 66

Figure 4.1 LigEvolver CompoundI 67

Figure4.2Similaritiesof compoundIwithknown inhibitors 68

Figure 4.3 LigEvolverCompoundII 68

Figure4.4 Similarities of compoundIIwithknown inhibitors 69

Figure4.5 LigEvolverCompound III 69

Figure4.6Similarities of compoundIIIwithknown inhibitors 70

List ofTables

Table 2.2.2. 1 Ligand

Summary

Table 2.3.1 ListofDatabases ₁₆

Table2.6.1.1.3.1 Amino Acidsandtheir_{SMILES string} 25

Table 2.6.3.1 Molecularweights 32

Table 2.6.4.1 Atomsandtheir_{Volumes 34}

Table 2.6.6.1 FDA

-Approved Protease Inhibitors 35

Table 2.6.7.1.1 Aromatic

RingCount _ScoringScheme 37

Table 2.6.7.2.1 Hydrogen Bond Donor Count_ScoringScheme 38

Table 2.6.7.3.1 _{Hydrogen Bond Acceptors CountScoringScheme 38}

Table 2.6.7.4.1 Hydrophobic Regions Count_ScoringScheme 38

Table 2.6.7.5.1 Molecular Weight_ScoringScheme 39

Table 2.6.7.6.1 Polar Surface Area_Scoring Scheme 39

Table 2.6.7.7.1 VanDer Waals Volume_ScoringScheme 40

Table 3.1.1 SimulationIProperties ₄₁

Table 3.2.1 Simulation II Properties 46

Table 3.3. 1 Simulation HI Properties 52

Table 3.4.1 SimulationIVProperties 58

Table 3.5.1 SimulationVProperties 61

Table 3.6.1 Simulation VI Properties 64

1.

Introduction

LigEvolver, the tool that produces novel ligand structures, has been implemented for

specific reasons. These intentions and the _waythis tool is different from previous genetic

algorithm based tools is presented in this section. The _necessary background and the

problem are alsodescribed.

1. 1 _{Fundamental Concepts}

Enzymes.

Enzymes are a specialized class of proteins that act as catalysts to increase the rate of

chemicalreactions.

Inhibitor

Theinhibitorsarechemical compoundsthatservetoblockenzymesfrom functioning.

Ligands

The substrate that is bound to _by a protein is also called a ligand. These can be used as

effectiveinhibitors. Theterms ligandandinhibitorareused_{interchangeably}inthis thesis.

GeneticAlgorithm

Genetic _{Algorithms employ} concepts of evolution to search for solutions in domain

problem spaces. The individualsin apopulation are made to crossover and mutateto give

apre-defined _{quality criterion,they} areeither allowed toexist or eliminated fromthe next

generation.

Pharmacophore

This is a common template that depicts the desired spatial arrangement of chemical

propertiesofa potential candidate. It istheminimal structure requiredforinhibition.

SMILES

This isan acronym for SimplifiedMolecularInputLine_Entry Specification. Itisa method

of_representing amoleculeinalinear form.

1.2 Purpose ofInvestigation

Inhibitordesign isa major part ofthe _{drug discovery}process. Structure-based_drug design

has been shown to produce effective results and has led to the _discovery of a number of

inhibitors. There are _plentyoffeatures thatlead to thedesignof a reliable _drug such as its

bioavailability,bindingaffinityandinteractionspecificity, amongotherfeatures.

In the case of enzymes with known _inhibitors, it would be worthwhile to explore the

possibility of _generating ligands based on scalar properties without other added

complexities. The focus of this _{study is directed} towards _determining if new ligand

structures could be derived by _using a minimal set of physiochemical properties and

Previous studiessuch as those_byGlenet_{al., (1994)} andVenkatasubramanianetal., (1995)

have shownthe useof_evolutionarytechniques for_drugdesign. Douguetet al made use of

SMILES line notation for their program LEA (Ligand _{by Evolutionary} Algorithm).

LigEvolver differs from the past research as it makes use of the smallest number of

constraints and genetic operators _{for evolving ligands. Its} main objective is to see if this

limited approach canleadtopowerful structures.

1.3 Problem Description

There are _many undiscovered therapeutic structures that have the _ability to function as

inhibitors. The goal of this _{study is} to uncover such structures _{by constructing} a genetic

algorithm based computer _tool, LigEvolver. The task is to _identify new inhibitors for the

HIV- 1 Protease

enzyme and LigEvolver aides this task _byevolving solutions that seem

more _likely to performthe task. The level of analysis of the chemical structures has been

restrictedto twodimensions becauseoftheminimalisthypothesis.

Given a set of chemical structures in the SMILES notation, LigEvolver evolves potential

inhibitors for the target enzyme. The algorithm conserves the fittest individuals in each

generation and iterates until _fairly competent candidates have been produced. Each

1.4 Rationale

Discovering new drugs is an _extremely tedious and expensive task. Computational

derivation ofpotential ligand structures would_help toreducethe time andthe costof_drug

development. The use of a genetic algorithm to generate candidate structures could give

riseto a_variety ofcombinations that wouldenable the scientist topursue lead compounds

thathavenot_{been identified using}traditionalapproaches.

The intention is to _develop a program _using a well-characterized system of an enzyme

(HIV protease) that has been crystallized in the presence of a large number on inhibitors.

In this case, the 3D structures of the drugs and the target protein are known. Once an

approachto_predictingnew ligandsbasedon _{existing ligands has beendeveloped,}itcanbe

extendedto open _cases, where a number ofdrugs havebeen identified,but thestructure of

2.

Methods

Materials

2. 1 Disease andEnzyme Selection

2.1.1 HIV Genome

The HIV genome consists of three main _{genes, namely gag,} pol and env. _Gag codes for

structural proteins that form the viral core. Pol genes code for Reverse Transcriptase, Integrase and _Protease, which arethreeessential enzymes that_play a significant roleinthe

formation and maturation of the virus. The env gene codes the envelope for the viral

proteins [39]. Figure 2.1.1.1 showstheHIVgenome_encompassingthe threegenes.

2.1.2 Lifecycle of HIV

Human_{immunodeficiency}virus_(HIV)is alethalretrovirus thatattackstheimmune

system.Thisretrovirus followsaspecificlife

-cycletoinfectthehostcells.Theviral

BINDING The HIVbinds itselftotheCD4

receptorinthecell membrane

FUSION

TRANSCRIPTION

INTEGRATION

Theviruspenetratesthe

membrane and entersthecell with

the_helpoftheco-receptor

Reverse Transcriptasetranscribes

theviralRNA into DNA

Integrase integratestheviralDNA

withthehostcell'sDNA

CLEAVING Proteasecleavestheviral

polyproteinsinto functional

proteins andthevirus exitsthecell

2.1.3 HIV Protease

Ascan beseen fromthe _{lifecycle, HIV Protease is} a crucial enzyme thathydrolyzes viral

polyproteins into functional protein products. These products then contribute to the

further propagate the virus. This work concentrates on _identifying new ligands for HIV

Protease. The_flexibilityand_mutabilityofthisenzymemakesit a_{challengingdrug}design

target. HIVprotease acts as adimerwithtwo identicalpolypeptidechains. _{It has only}one

active site thatis depicted intheboxedregionoffig. 2.1.3.1 [40].

Figure2.1.3.1 HIVProtease andactive site

*Figuregenerated_{using Deep}View

2.1.4 Conserved Region

During the transcription of RNA to _DNA, Reverse Transcriptase has an error rate of about 1 in 2000 bases. Dueto this the HIV strain undergoes mutation in each generation and the nucleotide sequence ofHIV Protease changes from strain to strain. _However, a conserved region has been detected. It isthe residues, Asp - Thr

-Gly, ofthe catalytic

triad.

Figure 2.1.4.1 Aspartic Acid(Asp)*

Figure 2.1.4.2 Threonine(Thr)*

)H H2N'

Figure 2.1.4.3 Glysine (Gly)*

O

"^p

1

H O _^

H3CT

1 II

L H O

OH

-R'

Figure 2.1.4.4 Catalytictriad

2.1.5 Enzymatic Reaction

The HIV protease enzyme cleaves the protein at the carbonyl _group ofthe peptide bond.

The entire process is shown with respect to the active site _(Asp - Thr

-Gly) in figure

2.1.5.1 [20]. _Step 1 shows the _binding ofHIV

-1 protease to a peptide substrate; _{step 2}

demonstratesthe attack ofthecarbonyl ofthescissile amide_byHIV - 1

proteaseactivated

water; step 3 illustrates the cleavage of the scissile bond while _step 4 depicts the final

H

\

I

K

/ / C=^ ff Asp25 .nr'% o^c/ (-) Asp25'

\

^

X7->c.

AsP^25 A -,r) Asp25'

C (-)

Xd

pr x-'"N< 1

H _{/ -,}

(-) H

9 / H

Asp25 Asp25'

^

4.

V

A

/ 9" h

\

H

A/

VH

c _\

Asp25

Asp25'

2.2 Genetic Representation in LigEvolver

2.2.1 Chromosome Representation

"Chromosome"

is the general term used to address the genotype of the individuals in a

population. _{The SMILES string} serves as the input to the genetic program. Its simplified

representation supplies thecomputer with an accuratelinearrepresentation ofcomplicated

two dimensional structures, which can be _readily manipulated _{during evolutionary}

programming [4]. It is also a standard and universal form of representation used _by

chemists all over the world [1]. These reasons make it a good choice for the depiction of

the chemical compounds and a means to represent the chromosome for the genetic

program. The _mapping of the chemical compound from its 3D structure to the SMILES

notationis shownin figure 2.2.1.1.

3D

2D

ID

CC(C)(C)NC(=0)ClCC2CCCCC2CNlCC(0)C(Cc3ccccc3)NC(-0)C(CC(N)=0)NC(=0)c4ccc5ccccc5n4

Figure 2.2.1.1 Transformation from 3D structureto_{SMILES string}

2.2.2 _{Sample SMILES string} and 2D structure of ligand

An example of the ligand structure is shown _{below along} with the SMILES string. The

common nameis_Tipranavir,whichis abbreviatedTPV.

N-(3-{(lR)-l-[(6R)-4-HYDROXY-2-OXO-6-PHENETHYL-6-PROPYL-5,6-DIHYDRO-2H-Name

PYRAN-3-YL]PROPYL}PHENYL)-5-(TRIFLUOROMETHYL)-2-PYRIDINESULFONAMIDE

TPV

HETID

Formula C3, H33 N2 05 F3S SMILES

CCCC2(cCclcccccl)CC(0)=C(C(CC)c3cccc(NS(=0)(=0)c4ccc(cn4)C(F)(F)F)c3)C(=0)02

String

T^\

/%

Figure 2.2.2.2TPVStructure

2.2.3 Pharmacophoric Features Considered

For each of the ligand structures the _following are the important features considered to

form the basic pharmacophore template. These features are thought to be important

interactionpointsbetweentheenzyme andtheligand.

1. AromaticRings [Ar]

2. HydrogenBond Acceptors [A]

3. HydrogenBondAcceptors andDonors _[A/D]

4. Hydrophobic center _[F]

A O

Figure 2.2.3.1 TPVStructurewithlabeledpharmacophoricfeatures

Key

Hydrophobic Center

A Hydrogen Bond Acceptor

A/D Hydrogen Bond Acceptor/Donor

Ar- Aromatic Ring

2.3Input

Before _beginningthe geneticalgorithm, certain pre

-processingstepshadto be completed

for the construction of the input files. SMILES databases, consisting of various chemical

compounds represented as SMILES strings, were assembled. The data for these databases

were extracted from various public databases. The public databases are listed in the

ListofDatabases

Kyoto EncyclopediaofGenes andGenomes _(KEGG)

ChemPDB

ChemBank

Macromolecular Structure Database _(MSD)

Anti - HIV National Cancer Institute

(NCI)

Protein Data Bank (PDB)

Table2.3.1 ListofDatabases

The above mentioned databases have been downloaded from Ligand.Info which is a

collectionof various small-moleculedatabases [27].

2.3.1 XML/Perl Parser

An XMLparser was constructedtoelucidatethedata fromthedatabases andtoproduce the

output in a format easily transferable to the database. In other words, the files from the

public databases contain information regarding a ligand such as its molecular _weight,

geometrical coordinates, etc. Theparser scans the_dump files forthetoken "SMILES" and

extracts _only the SMILES strings forthe compilation of the various input files. The lava

Parserworksthe same way.Dependingonthe type ofinputdata fileeither oftheparsersis

2.3.2 Protein Data Bank

A HIV database was constructed _{using the ligand data} of the various HIV

-1 Protease

inhibitors. These data were extracted from the Protein Data Bank _(PDB) and had to be

compiled _manually since the _{dump files from} the PDB did not contain _{SMILES string}

information of the ligands. This file contained the _identifier, molecular formula and the

SMILES stringinformation.

Public DBs r PDB Manual Input SMILES .txt HIVPDB .txt

$%

<*

Figure2.3.1 Input Construction

2.4 Output

Two files are produced as aresult of_running LigEvolver. Onegives the _{details regarding}

the number of children _produced, maximum, minimum and average fitness for each

generation. The otherfile gives the final set ofSMILES strings _along withthe generation

2.5 ExternalPackages

Certain calculations required the use of _PerlMol, an external package comprised of

modules for performing basic computational _chemistry related functions. This is freely

distributed software and a detailed description of PerlMol can be found at

(www.perlmol.org). PerlMolprovidedthe_followingfunctionalitiestoLigEvolver:

1. CalculationofChemical formula fromtheSMILES string

2. Calculationofbonds fromthe_{SMILES string}

3. PolarSurfaceArea_(PSA)

A perl script was written to combine the PSA script (available at

http://www.perlmol.org/examples/polar_surface_area/)withthechemicalformulaandbond

calculation.

2.6Design ofLigEvolver

This section gives adetaileddescription oftheindividual components ofLigEvolver. Each

LigEvolver

Genetic Algorithm

GeneticOperators

1

Fitness

Fragment Generator

Crossover

Mutator

Pharmacophore Elucidator

Molecular Weieht Calculator

Polar SurfaceArea

Van der Waals Volume

Motif Finder

Figure2.6.1 ComponentsofLigEvolver

2.6.1 Genetic Algorithm

A genetic algorithm _(GA) is used to evolve the population in a random manner to

generate_reasonablygoodindividuals. It is driven _bytheconceptofDarwinianevolution.

Theuse of genetic algorithmsinthegeneration ofligands has been successfullyexplored

[3,4,5]. The main parts of the algorithm fall into two broad categories _namely, Genetic

operators and Fitness function, which will be discussed in the _following sections. The

genetic algorithm is the main module. It comprises of functions to initialize the

population, mate individuals and generate new population. This is illustrated in fig.

Reset Matings Counter Decrement GenerationsCounter

Select Parents

Yes No

Generate Fragments

T

Mutate

Crossover

ProduceChildren

Yes No

Addtogene pool Discard fromgene pool

DecrementMatings

Theinitialpopulationisseeded in differentstyles.Theways ofinitialization are asfollows:

1. Initializewithknown HIVinhibitors

2. Initializewithrandominhibitors

3. Initializewitha mixtureofknownand randominhibitors

For each _generation, the population is rearranged _according to the fitness function and

only afixed number of individuals is retainedto contribute towards evolution while the

restarediscarded.

2.6.1.1 Genetic Operators

The genetic operators are themechanisms through which theindividuals in a population

are manipulated and changed. If evolution were viewed as a reaction, then genetic

operators can be considered catalysts. Two dominant operators observed in _many

applications are the Crossover and Mutation operator. _Depending on the need of the

problem customized operators _may also be included as is the case in this study. The

operators usedin LigEvolver are, GenFrag,CrossoverandMutator.

2.6.1.1.1 _GenFrag

The _{functionality} of this operator is to divide the _{input SMILES string into} meaningful

fragments. This operator was introduced because random segmentation ofthe SMILES

string maygive riseto _{chemically impossible}or meaninglessindividuals. Thenumbersin

Original _String

CC(C)CN(CC(0)C(Cclcccccl)

Fragmentsafter applicationof_GenFrag

The SMILES string isinitiallycutinto fragments.

<> <

CC(C)CN(CC(0)C(Cclcccccl)

< > < ?

Arandom pointofconcatenationischosentogive rise to twostringstoserveashead and

tail. Inthis casethe concatenation pointis taken tobetwo. So thestart ofthefragment is

combined with the second fragment for the head and the rest of the fragments are

combinedtogether toformthe tail.

Fragment One (Serves asthehead fortheCrossover_operator)

CC(C)CN(CC(0)C(Cclcccccl) NC(=0)OC2CCOC2)

Fragment Two (Servesasthe tailfortheCrossoveroperator)

(=0)c3ccc(N)cc3

Figure 2.6.1.1.1.1 Exampleof_GenFrag

Overlapand _duplicity checks areperformed withinthe _GenFrag,to avoid_{redundancy in}

the gene pool. The fragments to be combined for the head and the tail portion are

2.6.1.1.2 Crossover

The crossover operator is used to mate the parents. The _GenFrag operator returns two

fragmentsthataretheheadandtailofanindividual. Crossoverstakeplace90%ofthe time.

Thecrossover operator works as shownin figure 2.6.1.1.2.1.

SMILES

Parent 1

Head

CC(C)CN(CC(0)C(Cc1ccccd₎

Tail NC(=0)OC2CCOC2) SMILES Parent 2 Head CCC(c1 ccccd) Child 1 Tail C2=C(0)C3=C(CCCCCC3)OC2

CC(C)C(N)CN(CC(0)C(Cc1ccccd₎₎ C2=C(0)C3=C(CCCCCC3)OC2

Child2

CCC(d ccccd) NC(=0)OC2CCOC2)

Figure 2.6. 1.1.2.1ExampleofCrossover

2.6.1.1.3 Mutator

Mutations have played an important role in the evolution of _living organisms. These

operators cause variations in the gene pool. _They can be both beneficial and _harmful,

leads tono_adaptabilityin thepopulation whereas levelsof mutationsthataretoohigh lead

togeneticbreakdown as shown_{in fig. 2.6.1. 1.3[5].}

Off*- DNA

myiases -*On

Mutation rate

Evolutionary Immediate

death death

{no adaptability) (genetic _breakdown)

Figure 2.6.1.1.3 Mutationrates and genetic_{adaptability [5]}

In_LigEvolver,mutationstakeplace 10%ofthe time.

2.6.1.1.3.1 Amino Acid Insertion

Random places of insertion are chosen and an amino acid is included. The choice ofthe

Amino Acid SMILES _String

Alanine NC(C)C(0)=0

Arginine NC(=N)NCCCC(N)C(=0)0

Asparagine NC(=0)CC(N)C(=0)0

Asparatic Acid OC(=0)CC(N)C(=0)0

Cysteine NC(CS)C(=0)0

Glutamic Acid C(CC(=0)N)C(C(=0)0)N

Glutamine NC(CCC(=0)N)C(=0)0

Glycine C(C(=0)0)N

Histidine NC(Cc 1 cnc[nH]1)C(=0)0

Isoleucine CCC(C)C(N)C(=0)0

Leucine CC(C)CC(N)C(=0)0

Lysine NCCCCC(N)C(=0)0

Methionine C(N)(C(0)0)CCSC

Phenylalanine NC(Cclcccccl)C(=0)0

Proline C1CCNC1C(=0)0

Serine NC(CO)C(=0)0

Threonine CC(0)C(N)C(=0)0

Tryptophan NC(Cc1cc2ccccc2[nH]1 )C(=0)0

Tyrosine NC(Cc1ccc(0)cc1 )C(=0)0

Valine CC(C)C(N)C(=0)0

2.6.1.2 Parent Selection

Parent selection is another important design factor for a genetic algorithm. There are

various selectionalgorithms and theone employed inthis_{study is}tournament selection. In

thismethodtwogroupsoffour individuals areselectedrandomly.The one withthehighest

fitness score is selected fromeach group. These two individuals then serve as the parents.

In case of atie with respect to fitness _{score, the} parentischosen randomly. Thisselection

approachfollowstheconcept of atournamentwithteams_competingtoget selected. _Finally

thefitness scoreisthedeterminantofthewinners.

FinalistOne Finalist Two

Individual One IndividualThree Individual Four

2.6.2 Pharmacophore Elucidator

This module is used to extract the pharmacophore from a SMILES string. The

pharmacophoric features include aromatic _rings, hydrogen bond acceptor, hydrogen bond

acceptors/donors,hydrophobicregions,positivechargecenters and negative charge centers.

2.6.2.1 Aromatic Rings

Aromaticrings are a circulararrangement ofatoms thatcanberepresentedwith _alternating

single and double

-bonds. It is aconjugated system oflone pairs, unsaturatedbonds and

empty orbitals that put together form a _strong stabilization [36]. In chemical terms,

aromatic rings contain(4n+2)IIelectrons,where nisaninteger.

The _following arethe chemical compounds thatwouldbe detected_by the tool as aromatic

rings.

o

Figure2.6.2.1.1 Aromatic Structure I

- ABenzenering*

*

Drawinghas beengenerated_usingSmi2Depict developedby UniversityofCalifornia,

2.6.2.2HydrogenBondAcceptor

Hydrogenatomsthatarebound toelectronegative atoms such as oxygen or nitrogen can

interactwith lonepairstoformadditional_bonding [39]. The lonepairsthatinteractwith

thehydrogenatomaretermed ashydrogen bondacceptors.

The_following arethe chemical compoundsthatwouldbedetected _bythe toolashydrogen

bondacceptors.

Figure 2.6.2.2.1 Hydrogen Bond Acceptor Structure I

=

0

Figure 2.6.2.2.2 Hydrogen Bond Acceptor Structure II

O

N

Figure 2.6.2.2.4 Hydrogen Bond Acceptor StructureIV

2.6.2.3 Hydrogen Bond Acceptor/Donor

A groupthatcaneither acceptordonate hydrogen bonds is identifiedas hydrogenbond

acceptor/donor.

The_followingarethechemical compounds thatwould be detected_bythe tool as hydrogen

bond acceptors/donor. This is not an exhaustive _list, but contains a few of the common

bondsthatoccurinHIV inhibitors.

-OH

Figure 2.6.2.3.1 Hydrogen bond A/D Structure I

-NH2

Figure 2.6.2.3.3 Hydrogen bondA/D StructureIII

2.6.2.4 Hydrophobic Regions

Theregions thatrepel water aretermedhydrophobicregions.

The _following are the chemical compounds that would be detected _by the tool as

hydrophobicregions.

HaC ^ CHa

-N H

Figure 2.6.2.4.1 Hydrophobic StructureI*

-CHa

HaC

Figure2.6.2.4.2 Hydrophobic StructureIP

/s.

o

o

Figure 2.6.2.4.3 Hydrophobic Structure IIP

CH,

HaC-

//

o

Figure 2.6.2.4 .4Hydrophobic Structure

TV*

c

CH-,

CHn

c

Figure 2.6.2.4.6HydrophobicStructureVI*

A

O

Figure 2.6.2.4.6 Hydrophobic StructureVII*

*

Drawings have beengenerated_{using Smi2Depict developedby UniversityofCalifornia,}

Irvine _[44]

2.6.3 Molecular Weight Calculator

As the name _indicates, the molecular weight calculator computes the molecular weight

giventhe chemicalformula. The molecular weight values (round

-off_weight) used_bythe

toolis shown inthe tablebelow.

Molecule Weight Round- _{off weight}

C 12.0110369 12.011

H 1.0079759 1.007

B 10.8110280 10.811

N 14.0067231 14.006

Cl/1 35.4527379 35.452

Br/r 79.9035266 79.903

F 18.9984032 18.998

S 32.0643881 32.064

P 30.973762 30.973

I 126.904473 126.904

Fe/e 55.8451474 55.845

Na/a 22.9897677 22.989

Table 2.6.3.1 Molecularweights

2.6.4 Van DerWaalsVolume

Van derWaalssurfacesdefinetheclosest contactthatatomsina molecule could make with

one another. Hence it acts as anindicatorof atom_packing in a molecular structure. Italso

has animportantrolein _establishingthe _specificityofinteractions betweenprotein_binding

pockets andligands [36].

Zhao et alhave described a methodforthecalculation of vanderWaals volume_using the

formula, V(vdW) = _{summation operator all} atom contributions 5.92N(B) 14.7R(A)

3.8R(NR) (N(B) isthe numberof_bonds,R(A) is thenumber of aromaticrings, and_R(NA)

Table 2.6.4.1 givesthevalues oftheatoms andtheirvolumes.

Atoms _vdw

H 7.24

C _20.58

N 15.60

O 14.71

F 13.31

CI 22.45

Br 26.52

I 32.52

P 24.43

S 24.43

As 26.52

B 40.48

Si 38.79

Se 28.73

Te 36.62

2.6.5 Polar Surface Area

Polar Surface Area_(PSA) is a descriptorusedto indicate the _{bioavailability} ofdrugs. The

surface _belonging to polar atoms correlates with the transport of molecules through

membranes. Ertl et al proposed an approach for the calculation of PSA based on the

summation _{of surface contributions of polar fragments. PerlMol}

uses this method and

provides aperl scripttocalculatethePSA.

2.6.6 Motif Finder

Themotiffinder isresponsible_{for observingcommon patternsbetweena set ofstrings.It is}

used to derive structural similarities between compounds through the longest common

substringpresentintheirSMILES string.

The structures ofthe various FDA (U.S. Food and _{Drug Administration)} approved drugs

forHIVprotease were compiledintheformofSMILES strings. These werethen matched

withthe_recentlygeneratedligands todetermine if_theyhave acommon _substringoflength

greaterthan 9.Table2.5.6.1 isthelistofdrugsthathas beentakenintoconsideration.

Brand Name GenericName Manufacturer Name

Agenerase Amprenavir GlaxoSmithKline

Aptivus Tipranavir BoehringerIngelheim

Crixivan indinavir, IDV, MK-639 Merck

Fortovase Saquinavir Hoffmann-LaRoche

Kaletra lopinavirand ritonavir Abbott Laboratories

Lexiva FosamprenavirCalcium GlaxoSmithKline

Norvir _{ritonavir,ABT-538 Abbott Laboratories}

Reyataz Atazanavirsulfate Bristol-Myers Squibb

Viracept nelfinavir_{mesylate, NFV} Agouron Pharmaceuticals

Table 2.6.6.1 FDA

-Approved ProteaseInhibitors*

*Datatakenfrom FDAsite _[42]

2.6.7 Fitness Computor

The fitness function isusedtoevaluatethestrength and robustness of anindividualelement

in the population. The fitness is calculated for every entity and it determines if the

individual needs to beretained or eliminated.In a _{way, the}fitness function biases thepath

of evolution.The_followingisthefitness function.

F=Ar+Hd+Ha+_Hy+Mw +Ps+Vd+Sm

Where,

Ar- Aromatic

RingCountScore

Hd- Hydrogen Bond DonorScore

Ha- Hydrogen Bond Acceptor Score

Hy- Hydrophobic Region CountScore

Mw- MolecularWeight Score

Ps- PolarSurface AreaScore

Sm

-Structural Motif Score

For each of the above mentioned features scores were assigned _according to their

resemblance to the _{existing known inhibitors with the lowest score}

being the minimum

matching score and the highest _being the maximum match. Ifthere are no matches to the

known inhibitor feature count then no points are assigned. A list of the _{known existing}

inhibitors was made and theirfeature pattern was analyzed _manually to obtain the _scoring

scheme._{This list has beenincludedas appendixII.}

2.6.7.1 Aromatic

Ring

The Aromatic _ring is one of the pharmacophoric features that are vital components in

determining the _way the inhibitors interact with the enzyme. Table 2.6.7.1.1 gives the

scoringschemeforthearomatic_ringcount.

Aromatic _RingCount Score

>4and <7 1

>1 and<4 2

==4 ₃

Table 2.6.7.1.1 Aromatic_RingCount_ScoringScheme

2.6.7.2 Hydrogen Bond Donor Count

Acount ofthehydrogen bond donors istaken andscoresareassigned_accordingto table

Hydrogen Bond Donor Count Score

>1 and <5 ₁

>7 2

>4 and <8 ₃

Table 2.6.7.2.1 Hydrogen Bond Donor Count_ScoringScheme

2.6.7.3 HydrogenBond Acceptor Count

Acount ofthehydrogen bond acceptorsistakenandscoresare assigned_accordingto table

2.6.7.3.1.

Hydrogen BondAcceptorCount Score

>1 and<5 1

>7 2

>4and<8 3

Table 2.6.7.3.1 Hydrogen Bond Acceptors Count_Scoring Scheme

2.6.7.4 HydrophobicRegion Count

A countofthehydrophobicregionsistaken and scoresareassigned_accordingto table

2.6.7.4.1.

Hydrophobic RegionCount

>1 and<5

>7

>4 and <8 3

Table 2.6.7.4.1 Hydrophobic Regions Count_ScoringScheme

2.6.7.5 Molecular Weight

Scores are assigned _according to therange of molecular weights for each compound. The

range andtheir_{corresponding}scorearesummarizedintable2.6.7.5.1.

Molecular Weight _(Daltons) Score

<500 1

>600and<806 2

>500and<600 3

Table 2.6.7.5.1 Molecular Weight_ScoringScheme

2.6.7.6 Polar Surface Area

The ideal polar surface range has been takenfromthe list ofknown inhibitors and scores

are attributedto a_newlygeneratedindividualif itspolarsurface areafalls withintherange.

The details_regardingtherange andthe scores are givenintable2.6.7.6.1.

PolarSurface Area(Angstroms)

>150and<200

>100and<150

Score

2.6.7.7 Van DerWaals Volume

Van derwaals volume is another descriptorthat isused to assign scores to thepopulation

units.Thevolume range andtherespective scorehavebeenshownintable2.6.7.7.1.

Van Der Waals Volume(Angstroms3) Score

>400and<750 1

>800and<900 2

>900 and<1000 3

Table2.6.7.7.1 Van DerWaalsVolume_Scoring Scheme

2.6.7.8 Structural Motif

Thestructuresofthegovernment approveddrugswere comparedto thenew ligandtoseeif

they include similarities. The presence of a significant resemblance gives a score of one

3.

Results

LigEvolver was run _using the data from various public databases. The results obtained

are discussed inthis section. Unless otherwise specified theentire chemical figures have

been generated_{using CS ChemDraw Pro.}

3.1 Simulation I

Theinhibitorgenerationwas done usingtheparameters shownintable3.1.1.

Property Value

Database ChemPDB

InitialPopulation Size 4009

NumberofMatingsperGeneration 150

NumberofGenerations 100

Populationsize perGeneration 500

Table 3.1.1 Simulation I Properties

3.1.1 Significant Results

A few_noteworthy compoundsthatwere generated_duringthissimulationarediscussed in

3.1.1.1Generation#100

3.1.1.1.1 Compound1

SMILES string

C(=0)NC(Cc3ccccc3)C(NC(=0)C(N)c3ccccc3)C(=0)N2C

c2cccc3CC(C)CNc23

HoC

Figure 3.1.1.1.1.1 SimulationI- Generation 100- CompoundI*

3.1.1.1.1.2 Compound II

SMILES _String

NC(=0)clc[n+](cccl)C(NC(=0)C(N)c3ccccc3)C(=0)N2ClCCl=CN(C2)cccc3CC(C)C

Nc2cccc3CC(C)CNc23

Reconstruction

Elements underlined,eitherin the original _{SMILES string}or in thereconstructed string,

signify the portion ofthe _{SMILES string} that is modified. This convention is followed

throughout the_remainingsections.

NC(=0)c1c[n+](ccc1)C(NC(=0)C(N)c3ccccc3)C(=0)N2C1CCl=CN(C2)ccccCC(C)CN

c2cccc3CC(C)CNc23

^ /Nfl /N^

HoC

*Figuregenerated_{using ChemSketch [45].}

3.1.1.2Generation#85

3.1.1.2.1 CompoundI

SMILES_String

COC(=0)C(Cc1ccccc1)CCC1=C(C)C(=0)NC1

C(0)CN(CCC4CCC50COC5C4)C(=0)

CCN6C(=0)c7ccccc7C6

Figure 3.1.1.2.1.1 SimulationI-_{Generation 85}

3.1.1.2.2 CompoundII

SMILES _String

CC(C)C(NC(C)=0)C(=0)NC(Cc1ccccc1 )C(0)CN(CCC4CCC50COC5C4)C(=0)CCN6

C(=0)c7ccccc7C6

Figure3.1.1.2.2.1 SimulationI

3.2 Simulation II

Property Value

Database KEGG

Initial Population Size 10005

NumberofMatingsperGeneration 100

NumberofGenerations 45

Populationsize perGeneration 500

Table 3.2.1 Simulation II Properties

3.2.1 Significant Results

A few_noteworthycompounds thatwere generated_duringthis simulation arediscussedin

3.2.1.1 Generation #40

3.2.1.1.1 Compound 1

SMILES _String

CCC(C)C(N(C)C)C(-0)NClC(Oc2ccc(cc2)C(0)CNC(=0)C(Cc3ccccc3)NCl)CC(=0)N

C1C(0)CC(0C1)

Figure 3.2.1.1.1.1 Simulation II- Generation40

3.2.1.1.2Compound II

SMILES _String

CC(C)CC(NC(=0)C(NC(=0)NC(Cc1ccccc1)C(Oc4ccc(C=CNC(=0)C(Cc5ccccc5)NC3

=0)cc4)c6ccccc6

Reconstruction

CC(C)CC(NC(=0)C(NC(=0)NC(Cclcccccl))C(Oc4ccc(C=CNC(=0)C(Cc5ccccc5)NC=

0)cc4)c6ccccc6)

3.2.1.1.3 Compound III

SMILES _String

NC(=0)OC2CCOC2cccc3CC(C)CNc2cccc3CC(C)CNc2cccc3CC(C)CNc2cccc3CC(C)C

Nc23

Reconstruction

NC(=0)OC2CCOC2cccc3CC(C)CNc2cccc3CC(C)CNc2cccc3CC(C)CNc2cccc3CC(C)C

Nc2

Figure 3.2.1.1.3.1 SimulationII- Generation40- CompoundIIP

3.2.1.1.4 Compound IV

SMILES_String

CC(C)(C)NC(=0)C1 CN(CCN 1 )COC(=0)CC(0)(CCC(C)(C)0)C(=0)OC 1C2c3cc40C

Oc4cc3CCN5CCCC25C=C1

3.2.1.2 Generation #41

3.2.1.2.1 Compound I

SMILES _String

[nH]c(C=C4NC(=0)C(C)=C4)C(=0)N2CCCC20c2ccc(C=CNC(=0)C(Cc3ccccc3)NC(=

0)ClNC(=0)C4CCCN4C)cc2

Reconstruction

[nH]c(C=C4NC(=0)C(C)=C4)C(=0)N2CCCC20c2ccc(C=CNC(=0)C(Cc3ccccc3)NC(=

0)CNC(=0)C4CCCN4C)cc2

Figure 3.2.1.2.1.1 SimulationII- Generation41

-CompoundP

3.3 Simulation III

Property Value

Database MSD

Initial Population Size 6433

NumberofMatingsperGeneration 100

NumberofGenerations 100

Population sizeperGeneration 500

Table 3.3.1 SimulationmProperties

3.3.1 Significant Results

A fewnoteworthycompoundsthatwere generated_duringthis simulation arediscussedin

3.3.1.1 Generation #100

3.3.1.1.1 Compound1

SMILES _String

CN(C)C(C 1 CCNC1C(=0)Oc1ccccc1)C(=0)N2CCCC2C(=0)NC3C(Oc4ccc(C=CNC(= 0)C(Cc5ccccc5)NC3

Reconstruction

CN(C)C(C1CCNC1C(=0)0c1ccccc1)C(=0)N2CCCC2C(=0)NC3COcccc(C=CNC(=0)

C(Cc5ccccc5)NC3c4))

^^

Reconstruction

CN(C)C(C 1 CCNC1C(=0)Oc1ccccc1)C(=0)N2CCCC2C(=0)NC3C(Oc4ccc(C=CNC(= 0)C(Cc5ccccc5)N)C34)

Reconstruction

CN(C)C(ClCCNClC(=0)Oclcccccl)C(=0)N2CCCC2C(=0)NC3C(Oc4cccC=C4iNC(= 0)C(Cc5ccccc5)NC3

Figure3.3.1.1.1.3 Simulation HI- Generation 100- Compound I(c)*

*"

3.3.1.1.2 Compound II

SMILES _String

NC(CS)C(=0)OC2(0)C3C(OC)C1 (COC(=0)c6ccccc6)C45C(CCC8(C0C(=0)c6)C(=0)

NCC(=0)N(C)C(Cc2ccccc2)

Reconstruction

NC(CS)C(=0)OC2(0)C3C(OC)Cl(COC(=0)c6ccccc6)C23CCCC(COC(=0)cl)C(=0)N

CC(=0)N(C)C(Cc2ccccc2)

Figure 3.3.1.1.2.1 SimulationIII- Generation 100- Compound II

Reconstruction

NC(CS)C(=0)OC2(0)C3C(OC)Cl(COC(=0)c6ccccc6)C23C(CCClCOC(=0)c)C(=0)N

CC(=0)N(C)C(Cc2ccccc2)

\^<*

o

Figure 3.3.1.1.2.2 SimulationIII

-Generation 100- Compound II(b)*

*Figure has beengenerated_{using Smi2Depict developedby University}of_California,

Irvine [44].

3.3.1.1.3 Compound III

SMILES _String

CCCl=C(C)CN(C(=0)NCCc2ccc(cc2)S(=0)(=0)NC(=0)NC3CCC(C)CC3)C 1OCC1=C NC(=0)NClCc5cccnc5

3.4Simulation IV

Property Value

Database ChemBank

Initial Population Size 2344

NumberofMatingsperGeneration 100

NumberofGenerations 100

Populationsize perGeneration 300

Table 3.4.1 SimulationIV Properties

3.4.1 Significant Results

A few_noteworthycompoundsthatwere generated_duringthis simulation arediscussedin

3.4.1.1 Generation #100

3.4.1.1.1 Compound I

SMILES _String

NC(=0)C2CCCCN2C(=0)C(Cc3ccccc3)NC(=0)C(Cc4ccccc4)NC1 CN1CCN(CC 1)CC(

C)ClNC(=0)C(Cc2)

Reconstruction

NC(=0)C2CCCCN2C(=0)C(Cc3ccccc3)NC(=0)C(Cc4ccccc4)NC 1 CN1CCN(CC1_)CC( C)C1NC(=0)C(C)

3.4.1.1.2Compound II

SMILES _String

CN 1 CCN(CC1)CC(C)C 1NC(=0)C(Cc2ccccc2)NC(=0)C(CCCCCC(=0)NO)NC(=0)C3 CCCN3Clc5ccccc35

Reconstruction

CN1CCN(CC1)CC(C)C1NC(=0)C(Cc2ccccc2)NC(=0)C(CCCCCC(=0)NO)NC(=0)C3 CCCN3Clc5ccccc5

3.5 Simulation V

Property Value

Database HIV_PDB

Initial Population Size 71

NumberofMatingsperGeneration 100

NumberofGenerations 100

Population size perGeneration 150

Table 3.5.1 Simulation V Properties

3.5.1 Significant Results

Afew_noteworthycompoundsthatwere generated_duringthis simulationare discussedin

3.5.1.1 Generation #12

3.5.1.1.1 Compound I

SMILES_String

CC(C)(C)NC(=0)C1CC2CCCCC2CN 1 3=CN[C_@](C [C @_{H] (O)}[C@_H](Cc4ccccc4)NC(

=0)0[C@H]5CCOC5)(Cc6ccccc6)C3

Reconstruction

CC(C)(C)NC(=0)ClCC2CCCCC2CN13=CN[C](C[C@H](0)[C@H](Cc4ccccc4)NC(=

0)0[C@H]5CCOC5)(Cc6ccccc6)C3

3.5.1.2 Generation #10

3.5.1.2.1 Compound I

SMILES_String

CC 1CCC(0)C 1 CC(C)(C)NC(=0)C1CC2CCCCC2CN1C(0)C(0)C(OCc4ccccc4)C(=0)

NCc5c(F)cccc5

3.5.1.3Generation #9

3.5.1.3.1CompoundI

SMILES _String

(CNC(=0)c4ccccc3)CN(Cc2ccccc2)C(=0)COc3c(C)cccc3NC(=0)C(CC(N)=0)NC(=0)

c4ccc5ccccc5n4

Reconstruction

(CNC(=0)c4ccccc4)CN(Cc2ccccc2)C(=0)COc3c(C)cccc3NC(=0)C(CC(N)=0)NC(=0)

c4ccc5ccccc5n4

Figure 3.5.1.3.1.1 SimulationV

-Generation 9 - CompoundI*

*Figure has been generated_{using Smi2Depict developedbyUniversity}of_California,

Irvine [44],

3.6 Simulation VI

Property Value

Database HIVNCI

Initial PopulationSize 42689

NumberofMatingsperGeneration 100

PopulationsizeperGeneration 1000

Table 3.6.1 SimulationVI Properties

3.6.1 Significant Results

3.6.1.1 Generation #199

3.6.1.1.1 Compound I

SMILES _String

COC(=0)C(0)C(0)(CCC(C)C)C(=0)OClC2c3cc40COc4cc3CCN5CCCC25C=ClCCN

5CCCC25C=ClN=Nc5ccccc5

Reconstruction

COC(=0)C(0)C(0)(CCC(C)C)C(=0)OC1C2c3cc40COc4cc3CCN5CCCC25C=C1 CCN

5CCCC5C=CN=Nc5ccccc5

3.7_{Fitness Behavior}

It was noted that the fitness _steadily increased with the number of generations. After a

certain number ofgenerations the fitness values approached an upper limit and this value

remained constant through the subsequent generations. Figure 3.7.1 shows the fitness

behaviorchart across twosimulations.

AverageFitness

4. Discussion

The approachusedinthis_{study does} nottakeintoaccounttheentirecomplexitiesinvolved

in ligand design. It is limited due to the non - incorporation ofthe 3D features. The main

objective of this research was to investigate if scalar properties such as those used in the

fitness function would derive _promising ligand structures. Another aspect was to

experiment withtheapplicationof geneticalgorithmsto thedomain ofligandgeneration.

From the results it can be inferred that the tool generated ligands that were hopeful and

novel _by nature. It also proved that genetic algorithm is an efficient computational

approach towards _solvingproblems ofthis nature. Certain compounds that were produced

by LigEvolver _{had striking} similarities to _existing known inhibitors. A few of those are

presentedhere.

(a) 1NPW _(b) 1C6S

-CO

(c) 1W5X

Figure 4.2 SimilaritiesofCompoundIwithknown inhibitors[43]

\

o

l

UN'

>- HO-M.

y^,.

r

UH

\

V^

y

-rjH

c^-/Sr\.

Figure4.4Similaritiesof compoundIIwithknowninhibitor 1D4K[43]

?-,

(a)2A4F _(b) 1HWR

Figure 4.6 SimilaritiesofcompoundIIItoknown inhibitors [43]

Afew compounds were generatedthathada_totallynew structurethan theones_already

provedtobe inhibitors. Onesuch compoundfromthis classof resultis shownin Figure

4.7.

Figure 4.7LigEvolver CompoundIV*

LigEvolverCompound # InputDatabase

I Protein Data Bank_(onlyHIV_inhibitors)

II KEGG

III ChemBank

TV KEGG

5. Conclusion

LigEvolversucceeded_{in generating}new ligands usingtheminimalist approach. Theresults

have beenpromising, but given thenature oftheproblemthere is nocomputational _wayof

verifyingthesolutions.Aworthwhilefutureenhancement wouldbetoimprovisethefitness

function _{by incorporating} the 3D properties. Despite some unresolved _issues, this tool

provides an extensible _{first step} towards generation ofinhibitorsthatpossess thepotential

Bibliography

Paper References

1. David Weininger. _{(1987) SMILES,} a chemical language and information system.

1. Introduction to _Methodologyand _{EncodingRules. J. Chem. Inf. Comput. Sci.,}

28,31-36.

2. Schneider and Fechner. ₍₂₀₀₅₎ Computer - based de

novo design of_Drug - like

molecules. _{Nature, 4,}649- 663.

3. Glen and Payne. ₍₁₉₉₄₎ A genetic algorithm for the automated generation of

molecules within constraints. Journal ofComputer - Aided

molecular _{design, 9,}

181 202.

4. _Douguet, Thoreau and Grassy. ₍₂₀₀₀₎ A genetic algorithm for the automated

generation of small organic molecules: _Drug design _using an _evolutionary

algorithm. JournalofComputer- AidedMolecular

Design, 14,449- 466.

5. Genetic algorithm for the design of molecules withdesired properties. Journal of

Computer- Aided

molecular_design, 16(2002).

6. Miroslav Radman. (1999) Enzymes of _evolutionary change. _{Nature, 401,} 866

-869.

7. _Zhao, Abraham and Zissimos. ₍₂₀₀₃₎ FastCalculation of van der Waals volume

as asum of atomic andbondcontributions anditsapplicationto _drugcompounds.

JournalofOrganicChemistry, 68,7368- 7373.

8. Wlodawer and Vondrasek. ₍₁₉₉₈₎ Inhibitors of HIV - 1 Protease: A Major

Success of Structure - Assisted

Drug Design. Annual Review of Biophysical

9. Draghici and Potter. ₍₂₀₀₃₎

Predicting

Bioinformatics, 19,98- 107.

10. Tossi et al. ₍₂₀₀₀₎ Aspartic protease inhibitors: An integrated approach for the

design and synthesis of diaminodiol - based

peptidomimetics. Eur. J. Biochem,

267, 1715-1722.

11. Gohand Foster.EvolvingMolecules for

DrugDesign using Genetic Algorithms.

12. _You, Garwicz and _{Rognvaldsson. (2005) Comprehensive Bioinformatic Analysis}

ofthe _SpecificityofHuman _{Immunodeficiency}Virus Type 1 Protease. Journal of

Virology,79, 12477- 12486.

13. Fernandez et al. ₍₂₀₀₄₎ Inhibitor design_{by wrapping packing defects in HIV} - 1

proteins. _{PNAS, 101,} 11640- 11645.

14. Akaho et al. ₍₂₀₀₁₎ A _Study on _Docking Mode of HIV Protease and Their

Inhibitors. J. Chem._{Software, 7,} 103- 1 14.

15. Jones et al. ₍₁₉₉₇₎ Development and Validation of a Genetic Algorithm for

FlexibleDocking.J. Mol. Biol._267, 727- 748.

16. Head et al. ₍₁₉₉₅₎ VALIDATE: A New Method for the Receptor

-Based

Predictionof_BindingAffinitesofNovel Ligands. J. Am. Chem. _{Soc, 118,} 3959

-3969.

17. Griffith et al. _{(2004) Combining} structure

-based _drug design and

pharmacophores.Journal ofMolecularGraphics and_{Modelling, 23,}439- 446.

18. _Jones,WillettandGlen. ₍₁₉₉₅₎ Ageneticalgorithmfor flexiblemolecular_overlay

and pharmacophore elucidation. Journal ofComputer

-Aided Molecular Design,

19. Xue and Bajorath. _{(2000) Molecular Descriptors in Cheminformatics,}

Computational Combinatorial _Chemistry, and Virtual Screening. Combinatorial

ChemistryandHighThroughput_Screening,3, 363- 372.

20. Becketal._{(2002) Defining}HIV

-1 Protease Substrate Selectivity. Current_Drug

Targets

-Infectious_{Disorders, 2,} 37- 50.

Books

21. _Baker, B.R. (1967). Design _{of Active} - Site

-Directed Irreversible Enzyme

Inhibitors. New York: John_WileyandSons.

22. _{Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., Walter,} P. (2002).

Molecular_BiologyoftheCell.(FourthEdition).New York: Garland.

23. Gasteiger, J., & _{Engel, T.,} (Eds.). (2003). Chemoinformatics. New York: John

WileyandSons.

24. Mitchell,M. (1996).AnIntroduction toGenetic Algorithms. Cambridge,MA: The

MIT Press.

25. _Copeland, R. (2000).Enzymes:APractical Introductionto_{Structure, Mechanism,}

andData Analysis. (SecondEdition). NewYork: Wiley-VCH.

26. Kubinyi, H., & _Muller, G. ₍₂₀₀₄₎ Chemogenomics in _Drug Discovery: A

Medicinal_ChemistryPerspective. New York: John_WileyandSons.

Web References

27. Ligand.Info: Small-Molecule Meta-Database, http://ligand.info/, accessed as on

17th

28. Pharmacophore Identification and the Unknown Receptor Problem,

http://cnx.rice.edu/content/mll538/latest/, accessed as on

7th

Dec2005.

29. 3DPharmacophore_Searching,

http://www.netsci.org/Science/Cheminform/feature02.html, accessed as on

7th

Dec2005.

30. Survivalofthefittest in_DrugDesign,

http://pubs.acs.org/subscribe/iournals/mdd/v03/i09/html/felton.html,accessedas

on7thDec 2005.

31. Pharmacophore Perception _by Molecule Alignment,

http://www2.chemie.uni-erlangen.de/projects/sol/drugdesign.html,accessedason7thDec 2005.

32. GeneticAlgorithmsand_EvolutionaryComputation,

http://www.talkorigins.org/faqs/genalg/genalg.html, accessed as on

7l

Dec 2005.

33. SMILES Home Page, http://www.daylight.com/smiles/, accessed as on

7th Dec

2005.

34. California LutheranUniversity,

http://www.clunet.edu/BioDev/omm/hivrt/hivrt.htm,accessedas on28 Jan 2006.

35. California LutheranUniversity,

http://www.clunet.edu/BioDev/omm/hiv_protease/molmast.htm,accessedas on

28th

Jan 2006.

36. _Rutgers,Departmentof_ChemistryandChemicalBiology,

http://rutchem.rutgers.edu/content dynamic/faculty/edward arnold.shtml,

accessedason

28th

37. Research ofJoanna_{Trylska, http://www.trvlska.com/research.html.}accessedas on

28th

Jan 2006.

38. _{Wikipedia, http://en.wikipedia.org/wiki/. accessed ason}28th

Jan 2006.

39. Themoleculesof_{HIV,http://www.mcld.co.uk/hiv/?q=HIV%20genome,}accessed

as on 16th_May 2006.

40. _BiologyofHIV Protease_Inhibitors,

www.unc.edu/~brybar/evolutionmodule/Biologv%20of%20HIV%20Protease%20

Inhibitors.doc. accessed ason

16th

May2006.

41. UniversityofCalgary,

http://www.chem.ucalgary.ca/courses/351/Carey5th/Ch27/ch27-4-4.html

accessed on

10th

March 2006.

42. FDAApproved_Drugs,http://www.fda.gov/oashi/aids/virals.html,accessedas on

10th

March 2006.

43. Protein Data_Bank,http://www.rcsb.org/pdb/home/home.do,accessed ason 12th

October 2005.

44. ChemDB /Smi2Depict:Generate 2D Images fromMoleculeFiles,

http://cdb.ics.uci.edu/CHEM/Web/cgibin/Smi2DepictWeb.py, accessed as on 15th

December2005.

45. ACD/ChemSketch 10.0_Freeware,

http://www.acdlabs.com/download/chemsk.html, accessedason 18th

November,

Appendix I

SmilesEncodingRules

1. Non

-hydrogen atoms are specified _by their atomic symbol enclosed in square

brackets. The second letter of the two - letter

symbol must be entered in lower

case.

2. Elements inthe "organicsubset",B, C, N, O, P, S, F, CI,Brand_{I,may be} written

without brackets if the number of attached hydrogens conforms to the lowest

normal valence consistentwith explicitbonds.

3. Elements notintheorganic subset mustbe described in brackets.

4. Attached hydrogens andformal charges are always specifiedinside brackets. The

numberof attachedhydrogens is shown_bythe symbolHfollowed _byan optional

digit_{(e.g., [NH4+]}

-ammoniumcation).

5. _Similarly, aformal chargeis shown_byone ofthe symbols +or-, followedby an

optional digit (e.g., [Fe+2] - iron

(II) cation). The form _[Fe+++] is considered

synonymous withtheform[Fe+3].

6. Ifunspecified, the number of attached hydrogens and charges is assumed to be

zeroforan atominsidethebracket.

7. Atoms inaromatic rings are specified_bylowercase_letters; e.g.,normal carbon is

represented_bytheletterC,aromatic carbon_byc.

8. Singlebond isrepresented_bythesymbol

-9. Double bondisrepresented_bythe symbol=

10. Triplebond isrepresented_bythesymbol#

12. Singleand aromaticbonds may_be, and_usuallyare, omitted.

13. Hydrogenscan beomittedfrom theSMILESnotation

Structure

CH2=CH-CH2-CH=CH-CH2-OH

Valid SMILES

c=ccc=cco

c=c-c-c=c-c-o

occ=ccc=c

14. Branchesare enclosed withinparentheses

Triothylamine

CH3

I

CH2

I

H2C CH2 N CH2 CH3

SMILES CCN(CC)CC

Isobutyricacid

CH3 O

I

II

H3C CH C OH

SMILES CC(C)C(=0)0

3

-propyl- 4

-isopropyl- 1

-heptene CH3 CH2

I