Arora R. (ed.) Medicinal Plant Biotechnology

Подождите немного. Документ загружается.

©CAB International 2010. Medicinal Plant Biotechnology 244

(ed. Rajesh Arora)

Chapter 15

Docking-based Virtual Screening of Anticancer

Drugs

Feroz Khan, Abha Meena and Ashok Sharma

Introduction

To combat life-threatening diseases such as AIDS, cancer, tuberculosis, malaria etc., the

only way to fight is by the development of new, effective drugs, at a fast pace. Drug

discovery and development is an intense, lengthy and an interdisciplinary endeavour. Drug

discovery is mostly portrayed as a linear, consecutive process that starts with target and

lead discovery, followed by lead optimization and pre-clinical in vitro and in vivo studies to

determine if such compounds satisfy a number of pre-set criteria for initiating clinical

development. For the pharmaceutical industry, the number of years to bring a drug from

discovery to market is approximately 13–15 years and costing up to $1.3–$1.5 billion

dollars. Traditionally, drugs were discovered by synthesizing compounds in a time-

consuming multi-step process against a battery of in vivo biological screens and further

investigating the promising candidates for their pharmacokinetic properties, metabolism

and potential toxicity. Such a development process has resulted in high attrition rates with

failures attributed to poor pharmacokinetics, lack of efficacy, animal toxicity, adverse

effects in humans and various commercial and miscellaneous factors. In addition to that,

toxicity/dosage in the form of IC

(inhibitory concentration) or LD

(lethal dose) of drugs

is determined by testing on laboratory animals, unnecessarily killing them. Moreover, if

multiple drug resistance is developed against any of these drugs that required so many

years to be developed, the whole process would have to be started all over again, where

again we are left without a drug against the disease. Along with the current problems, once

a biologically active compound has been synthesized in the laboratory not all are suitable to

be used as drugs due to toxicity, unfavourable side-reactions, or pharmacokinetics. After

the synthesis and aside from the desirable therapeutic properties, the testing and approval

of many rigorous tests are required to ascertain whether the compounds are worthy of

becoming drugs. An undesirable proportion of compounds with good biological activity

fails to progress to later stages of drug development because of inappropriate

pharmacokinetic and pharmacodynamic properties. Today, the process of drug discovery

has been revolutionized with the advent of genomics, proteomics, bioinformatics and

efficient technologies such as combinatorial chemistry, high throughput screening (HTS),

Virtual Screening of Anticancer Drugs 245

virtual screening, de novo design, in vitro, in silico ADMET screening and structure-based

drug design. Computational models are useful because they rationalize a large number of

experimental observations and therefore allow saving time and money in the drug design

process. In addition, they are useful in areas like design of virtual compound collections,

computational-chemical optimization of compounds and design of combinatorial libraries

with appropriate ADME properties.

An increasing number of protein crystallographic structures are becoming available

from high-throughput structural genomics projects, thus prediction of a potential lead and

its potential target is a fundamental step in order to investigate the molecular recognition

mechanism of a protein (Abagyan et al., 1994; Abagyan and Totrov, 2001). For lead

identification and optimization, wet lab and dry lab (in silico) methods are effectively

applied to speed-up the process of drug discovery (Anderson and Chiplin, 2002). The lead

molecules modulate the function of the target proteins and are later optimized to

therapeutic drug against a specific disease (Gane and Dean, 2000; Klebe, 2000).

Nowadays, to check the binding affinity of the target receptor with the library compounds

computational screening method e.g. Virtual Screening (VS) is widely applied and used by

many researchers to save the wet lab economy and time (Walters et al., 1998;

Subramaniam et al., 2008). Here we attempt to virtually screen the chemical compounds

and phytomolecules against known anti-cancer targets based on both Quantitative

structure-property/activity relationships (QSPR) and docking studies without any prior

knowledge about the nature of the interaction and target’s active sites (Shoichet et al.,

2002; Veber et al., 2002). We postulate that the structure itself is informative enough, thus

consider the whole protein for docking studies. To generate a list of potential drug

candidates, drug likeness was established in parallel to high throughput ligand docking

(Lipinski, 2000). We have demonstrated that our approach can successfully identify and

predict leads and their binding targets in both complexed and apo-structures (Klebe, 2000).

For large-scale identification, we used QSPR and docking parameters on diverse

compounds e.g. antimalarials, antipruritics, antifungal, antiprotozoal, anti-infective,

coccidiostats etc. and are able to generate good results. This fast and accurate method can

be used to identify possible binding for orphan receptors and vice versa, or for

uncharacterized secondary binding of known receptors. In the studied work, QSPR,

Lipinski’s rule of five filter and docking studies were performed for the virtual screening of

potential leads, suggesting action of compounds on the anticancer targets. A total of 100

compounds were used for library building, mostly belongs to anticarcinogenic

pharmacological category. For automated docking strategies, we used 16 known anticancer

targets. Results showed that only 68 compounds were found eligible as potential anticancer

drug leads. This fast and accurate method can be used to identify possible binding for

orphan receptors, or for uncharacterized secondary binding of known receptors. It can also

be used to prioritize novel targets.

Methodology

Structure cleaning and preparation

Crystal structures of anticancer receptors obtained from the Brookhaven Protein Databank

(Berman et al., 2000) were converted from .pdb to .csf format using standard functionality

from the Workspace module of Scigress Explorer 7.7.0.47. This included automatic typing

of each ligand atom based on the geometry found in the crystal structure and addition of

246 Virtual Screening of Anticancer Drugs

hydrogens as necessary. Atom typing of protein residues and addition of hydrogens to them

was carried out by reference to standard protein libraries keyed to residue type. Terminal

rotors such as methyl and hydroxyl groups were then relaxed to avoid distortion of scores

by spurious steric clashes with the added hydrogen atoms. To ensure the compound

availability or the synthetic feasibility, we used only reported anticarcinogenic compounds

for VS. We also analysed and consider tautomeric states of molecules for screening, since

these shows interaction with the active site residues of the target (Lipinski et al., 1997).

Collecting uncomplexed structures

To find out the potential active site with minimum docking interaction energy in the

uncomplexed protein structure (apo-structures), we consider cleaned monomer unit of

target protein as an active site for docking based virtual screening. Note that unliganded

pockets may not be as obvious as the liganded pockets due to the ligand induced

conformational changes. Side-chains often obstruct a part of the pocket in the absence of

the ligand. Scigress Explorer docking runs utilized the default parameters provided with the

docking module.

Building of virtual compound/phytomolecules library

We built an in silico compounds and phytomolecules library prior to virtual screening

based on anticarcinogenic activity. All the compounds and phytomolecules were retrieved

through PubChem database at NCBI web server. To made the diversified compound library

and also for comparative study, we randomly included compounds and phytomolecules in

the library from other pharmacological categories. For comparison we grouped all the

compounds in three datasets viz.: (i) known or reference data set, which includes only

reported leads e.g. allicin, catechin and camptothecin against studied 16 anticancer

receptors; (ii) unknown data set, which includes all the anticarcinogenic compounds and

phytomolecules without any information of receptors/targets; and (iii) diverse unknown

data set, which includes compounds/phytomolecules other than anticarcinogenic activity

e.g. anti-malarial. The basis behind such an analysis is that, similar compounds exhibit

similar physicochemical and biological properties. A chemical library enriched with the

representatives of dissimilar compounds or diverse chemical structures shall increase the

probability of finding different leads with similar biological activities. It also ensures that

the library is small but diverse enough to be readily managed.

Descriptors for small molecules

The problem of explicitly representing and storing small molecules as finite-dimensional

vectors has a long history in chemoinformatics and a multitude of molecular descriptors

have been proposed. These descriptors include in particular physicochemical properties of

the molecules, such as its solubility or logP, descriptors derived from the 2D structure of

the molecule, such as fragment counts or structural fingerprints, or descriptors extracted

from the 3D structure. We evaluated the performance of all combinations of these

descriptions on the compound library data, which contains three known and 86 potential

anticarcinogenic lead molecules and 11 unknown diverse data. However, in all cases, we

obtained significant improvements of the screening accuracy with respect to the individual

setting.

Virtual Screening of Anticancer Drugs 247

Virtual screening of library

In silico virtual screening of receptors is however a daunting task, both for receptor-based

approaches (docking) and for ligand-based approaches (QSPR/QSAR). The former relies

on the prior knowledge of the 3D structure of the protein, while the latter relies on the prior

knowledge of the chemical’s structure property and activity relationships. However, their

accuracy is fundamentally limited by the amount of known ligands and degrades when few

ligands are known. Docking studies includes two goals: to rapidly determine the most

likely binding mode of leads and to return some measure of its estimated binding affinity

for the target protein. In the studied work, we used a combination of both receptor and

ligand based screening methods. Firstly we used cleaned 3D structures of the known and

potential anticancer targets to search for potential phytomolecules or compounds that can

modulate the receptor function. Each of the library compounds is docked into the known

anticancer receptor binding site and the best one is predicted. Secondly we used

physicochemical properties of the anticarcinogenic compounds and phytomolecules for

screening potential leads based on QSPR and docking studies. For the drug likeness filter

we used the ‘Rule of Five’ screening method. Finally potential leads are ranked on the

basis of QSPR and docking scoring function. During molecular docking, we avoided target

flexibility parameter and use of homology models since we included the reported

anticancer receptors. A brief sketch of methodology has been shown in Fig. 15.1.

Drug-likeness filter

The drug-like compounds represent adsorption, distribution, metabolism and excretion

profiles as determined by Lipinski’s rule of five. This method is not a QSTR as there is no

quantitative structure–toxicity relationship with any experimental data. The number of

‘violations’ of Lipinski’s rules is claimed to be an indication of the absorbability of a

molecule. The drug-like compounds resemble the existing drug molecules and show

properties such as MW< 500, LogP <5, H bond acceptors <10 and H bond donors <5.

Toxicity filter

To minimize expensive drug failures in the later stage of drug discovery, toxicity related

experimental data are essential. Prediction of toxicity before the synthesis of compounds

ensures the removal of compounds with potential toxic effects. Considering this fact, we

used only those compounds which were experimentally verified in prior toxicity studies.

Compound’s safety and toxicological data were verified through authentic toxicity

databases especially for anticarcinogenic compounds such as CCRIS– carcinogenicity,

tumour promotion, tumour inhibition and mutagenicity tests (http://toxnet.nlm.nih.gov),

EINECS – European Inventory of Existing Commercial Chemical Substances

(http://ecb.jrc.it/esis/), GENETOX– genetic toxicology information

(http://toxnet.nlm.nih.gov/), TOXLINE– citations to the toxicological literature

(http://toxnet.nlm.nih.gov/) and NTP DBS– toxicological assay results (http://ntp-

apps.niehs.nih.gov/ntp_tox/).

248 Virtual Screening of Anticancer Drugs

Fig. 15.1. Methodology flowchart for virtual screening of anticancer leads and

development of QSAR/QSPR model.

Binding affinity of proteins/ligand

Estimation of ligand–protein affinity/ligand–DNA or drug–target affinity is one of the

major and most basic steps in drug discovery, developing QSAR models etc., since only

those potential molecules are taken up further for analysis which demonstrate a desirable

binding for the targeted receptor. This technique also helps in determining the

pharmacological activity of the lead compound and thus develops its QSAR models.

Binding affinity is calculated by docking, based on different kinds of scoring functions

such as PMF (potential of mean force), PLP, G-score, F-score, LigScore, X-score etc.; the

basic assumption here is that a good docking score indicates that the molecules bind to each

other.

Library compounds

(100)

Library compounds

violating the Rule of

Five

(29)

Library compounds

favouring the Rule of

Five

(71)

Good

docking

compds.

(

)

Poor

docking

compds.

(

)

Good

docking

compds. (68)

Poor docking

compds. (3)

Potential

leads

Reference

set (known

Test set

(library

QSAR/QSPR/QSTR model development for

the prediction of activity/property/ toxicity

data and responsible descriptors.

Virtual Screening of Anticancer Drugs 249

Quantitative Structure Activity/Property Relationship (QSAR/QSPR)

Structure–function correlation studies aim at broadening the understanding of relationships

between molecular intrinsic chemical features and biological properties. While many

QSAR analyses involve the interactions of a family of molecules with an enzyme or

receptor binding site, QSAR can also be used to study the interactions between the

structural domains of proteins. Protein–protein interactions can be quantitatively analysed

for structural variations resulted from site-directed mutagenesis. In simple words QSAR

help us in understanding how a particular property e.g. high melting point or a

pharmacological property such as anti-inflammatory might be attributed to a side chain,

chemical group, ring structure etc.

For example, see Table 15.11, summarizing an anti-malarial QSAR model for the

prediction of biological activity (IC50) and other dependable chemical structure’s

descriptors.

QSAR Model Equation:

Predicted log IC

(D) = - 0.071944*F + 0.00188086*H + 0.0844403*I + 0.650836*K +

0.0467254*T + 0.598326*AA - 0.306307*AD + 1.45219

Where, F= Atom Count (all atoms), H= Conformation Minimum Energy (kcal/mol), I=

Dipole Moment (debye), K= Dielectric Energy (kcal/mol), T= Polarizability, AA= Ring

Count (aromatic 6 membered) and AD= Size of Smallest Ring. With best fit relationship

regression line showing regression value of r

=0.998482 and prediction accuracy of

rCV

=0.994135, where, r

= Regression coefficient and rCV

= Cross-validated regression

coefficient. (Standard error = 0.029061)

Mostly it would be better to consider following classes of chemical structure

descriptors, while developing QSAR model:

1. Constitutional descriptors: molecular weight, halogenated atom counts and functional

group counts (amine, carbonyl, carboxylic acid, hydroxyl, nitrile, nitro group) etc.

2. Electrostatic descriptors: partial charges, charged surface areas etc.

3. Topological descriptors: Kier and Hall connectivity indices (zero order, first order and

second order) and valence connectivity indices (zero order, first order and second order)

etc.

4. Geometrical descriptors: solvent accessible surface, isosurface volume, molar volume

etc.

5. Quantum-chemical descriptors: dipole moment, quadrupole moment, polarizibility,

HOMO and LUMO energies, dielectric energy etc.

For example, LogP: The octanol-water partition coefficient. Solvent accessible surf area

(SASA): The molecular surface area accessible to a solvent molecule. Polarizability (P):

The molecule's average alpha polarizability. Smallest ring size (SRS): Information about

rings present in the compound. Rings with more than 12 atoms are ignored. The number of

atoms forming the smallest ring in the compound, or 0 if the compound contains no ring of

size 12 or less. Ring count (RC): The number of rings present in the compound. Rings

with more than 12 atoms are ignored. The number of rings with 12 or fewer atoms (All= all

250 Virtual Screening of Anticancer Drugs

aromatic, small, 5-membered, 5-membered aromatic, 6-membered, 6-membered aromatic,

large, large aromatic). Molar refractivity (MR): The molar refractivity of the compound.

LUMO energy: The energy gained when an electron is added to the lowest unoccupied

molecular orbital (LUMO). Ionization potential (IP): The energy required to remove an

electron from a molecule in its ground state. HOMO energy: The energy required to

remove an electron from the highest occupied molecular orbital (HOMO). Heat of

formation (HF): The energy released or used when a molecule is formed from elements in

their standard states. Conformation minimum energy (CME): Energy calculated for an

optimized conformation of the compound. Formal Charge (FC): The net positive or

negative charge on the molecule. Atom count (AC): The number of atoms. Bond count

(BC): The number of bonds. Weak and ionic bonds are ignored.

Docking

As mentioned above, for high throughput screening of the lead compounds, whether using

the target based or ligand based approaches, we need to dock the ligand into the receptor.

Essentially docking is a method which predicts the preferred orientation of one molecule to

a second when bound to each other to form a stable complex. Knowledge of the preferred

orientation in turn may be used to predict the strength of association or binding affinity

between two molecules using scoring functions. Molecular docking can be thought of as a

problem of lock-and-key, where one is interested in finding the correct relative orientation

of the key which will open up the lock. Here, the protein can be thought of as the ‘lock’

and the ligand can be thought of as a ‘key’. Molecular docking may be defined as an

optimization problem, which would describe the ‘best-fit’ orientation of a ligand that binds

to a particular protein of interest.

Results and Discussion

QSPR approach was used to develop a regression equation that describes the correlation

between experimental data and certain calculated descriptors based on the molecular

structure. The derived equation was then used to predict the properties of new or

hypothetical molecules from their structure alone, reducing the need for experimental

synthesis and measurement. The purpose of studied work was to find the best correlation

coefficient (r

) and the best cross-validated coefficient (rCV

), between the experimental

data and various descriptors that we calculated, while minimizing the number of descriptors

used. Correlation coefficient indicates how well the experimental data fits the regression

equation, while cross-validated coefficient indicates the ability of the regression equation to

predict unknowns. Results indicate values closer to unity, thus showed better correlation.

The ‘Lipinski’s Rule of Five’ predicts that poor adsorption or permeation of a drug is more

likely when there are more than 5 H-bond donors, more than 10 H-bond acceptors, the

molecular weight is greater than 500, and the calculated LogP is greater than 5. Molecular

weight and LogP can be calculated directly. The total number of hydrogen atoms bonded to

oxygen (OH) or nitrogen (NH) is considered to be a good index of H-bond-donor count,

while the sum of the oxygen atoms and nitrogen atoms is considered to be an acceptable

indicator of H bond acceptor count.

Virtual Screening of Anticancer Drugs 251

Our results demonstrate that VS approaches outperform individual approach, in

particular in cases where very limited or no ligand information is available. In the case of

well studied targets, more classical ligand-based methods (QSAR) may be better suited to

predict new strong binders from a large number of known ligands. Since normally VS

approach is a ligand-based approach, it would probably be interesting to use it in

combination with docking. Indeed, although prior known ligands can help tuning docking

procedures to the receptor under study, it can in principle be used with little or no ligand

information. When more experimental 3D structures become available for Targets in the

future, this will help building reliable models for a wider range of Targets that would be

suitable for docking studies. Joint use of ligand-based VS and docking would certainly

improve predictions.

Instead of focusing on each individual target independently from other proteins, we

used VS to screen compounds against several targets of the same family simultaneously.

This systematic screening of interactions between the compounds and protein targets can

be thought of as an attempt to fill a large two dimentional interaction matrix, where rows

correspond to compounds and columns to protein targets. Compound entry in the row

indicates whether it can bind to target in the column or not. While in general the matrix

may contain some description of the strength of the interaction, such as the docking score

of the complex, which reflect minimum binding interaction energy. While classical docking

or ligand-based virtual screening strategies focus on each single row independently from

the others in this matrix, i.e., treat each target independently from each others, the VS

approach is motivated by the observation that similar molecules can bind similar proteins,

and that information about a known interaction between a ligand and a target could

therefore be a useful hint to predict interaction between similar molecules and similar

targets. This can be of particular interest when, for example, a particular target has few or

no known ligands, but similar proteins have many: in that case it is tempting to use the

information about the known ligands of similar proteins for a ligand-based virtual

screening of the target of interest. In this context, we can formally define VS as the

problem of predicting interactions between a molecule and a ligand (i.e., the unknown

entries of the matrix) from the knowledge of all other known interactions or non-

interactions (i.e., the known entries of the matrix refer by reference data set).

Example: Case study on anticancer compounds

In the studied work, all the 100 compounds of library were used for QSPR relationship

comparison, rule of five filtering and docking studies separately. For the identification of

potential leads, we filter out compounds violating the ‘Rule of Five’ for drug likeness. In

this case study example, 71 out of 100, library compounds follow the Lipinski’s rule of five

(Table 15.1), while rest 29 compounds violate the Lipinski’s rule, thus rejected as potential

leads (Table 15.2). These 29 compounds predicted to have poor adsorption or permeation

ability because of one or more following features: more than 5 H-bond donors, more than

10 H-bond acceptors, the molecular weight is greater than 500 and the calculated LogP is

greater than 5. Out of these 29 compounds, 11 showed poor interaction energy with known

16 anticancer receptors through molecular docking studies (Table 15.3), while 18 showed

good interaction energy (Table 15.4) with known 16 anticancer Human and mammalian

receptors (Table 15.5), this indicates that apart from rule of five, some compounds

exceptionally may act as good leads, but may show low efficacy power in terms of

pharmacokinetic studies. On the other hand, 68 out of 71 compounds showed both good

252 Virtual Screening of Anticancer Drugs

docking interaction with known anticancer receptors and also favours the Lipinski’s rule of

five, thus selected as potential anticancer leads (Table 15.6), while 3 compounds showed

poor docking interaction with receptors, thus rejected for further studies. After docking

studies, quantitative data prediction analysis have been performed for best selected 68

compounds based on 3D chemical structures (Table 15.7), so that to built the QSAR/QSPR

mathematical model for the prediction of biological activity of unknown compounds and

also to find out the dependable descriptors responsible for anticancer activity through

regression statistical analysis (see example of anti-malarial QSAR model in Table 15.11).

Table 15.1. The library compounds (71) favours the ‘Lipinski’s Rule of Five’ as a test for

drug likeness or druggability.

PubChem Compound ID (71 in numbers) Druggability Test

CID_16741, CID_47318, CID_65036, CID_3712,

CID_5280373, SID_12013078, SID_210967, SID_49834379,

SID_17396662, CID_24360, CID_16590, CID_11617,

CID_9685, CID_5708, CID_5350, CID_5311, CID_5057,

CID_5056, CID_65781, CID_3752, CID_4018, CID_10630736,

CID_10678455, CID_10923720, CID_11000304,

CID_11778794, CID_11821800, CID_23663982, CID_68781,

CID_114835, CID_114866, CID_147004, CID_440623,

CID_495574, CID_634470, CID_638034, CID_1548910,

CID_3001662, CID_3070413, CID_4141487, CID_5281232,

CID_5282435, CID_5381733, CID_5385552, CID_5472495,

CID_6158261, CID_9064, CID_969516, CID_5281767,

CID_442783, CID_10666836, CID_10595440, CID_16666,

CID_1254, CID_165675, CID_19244, CID_10607, CID_72,

CID_3071, CID_6398761, CID_6712212, CID_9547315,

CID_9776851, CID_9802152, CID_9931851, CID_10084664,

CID_10251785, CID_10378578, CID_10523508,

CID_10534893, CID_10612216

Lipinski’s Rule of Five:

- Mol. weight < 500

- LogP < 5

- H-bond donors < 5

- H-bond acceptors < 10

Inference =

Drug like properties

Table 15.2. The library compounds (29) violating the ‘Lipinski’s Rule of Five’.

PubChem Compound ID (29 in No.) Druggability Test

CID_16219940, CID_16197727, CID_122127, CID_65064,

CID_446925, CID_5978, CID_5385555, CID_13342,

CID_5280531, CID_5319123, CID_5316150, CID_5281235,

CID_72495, CID_10794036, CID_5288209, CID_10532024,

CID_401405, CID_10428109, CID_23639625, CID_10337961,

CID_82146, CID_21773740, CID_9960816, CID_11969852,

CID_9847391, CID_9832869, CID_9830646, CID_65076,

CID_4259412

Lipinski’s Rule of Five

violation:

- Mol. Weight > 500

- LogP > 5

- H-bond donors > 5

- H-bond acceptors > 10

Inference =

Poor permeability/

absorption through cell

membrane

Virtual Screening of Anticancer Drugs 253

No.

PubChemID Docking score (in Kcal/mol) against known

Anticancer drug targets

(PDB ID: 1a25, 1a35, 1agw, 1awn, 1cdk, 1cpj, 1crp,

1dvi, 1jsu, 1mnc, 1p38, 1rpa)

1. CID_11969852 10.2

2. CID_5385555 4586.9

3. CID_446925 13605.4

4. CID_5319123 13686.0

5. CID_5978 37283.7

6. CID_5316150 62019.1

7. CID_5281235 75635.4

8. CID_13342 76808.1

9. CID_16219940 979421.3

10. CID_122127 1118907.0

11. CID_16197727 12333424.0

Table 15.4. The library compounds violating the Lipinski’s Rule of Five and showing good

docking interaction energy (in Kcal/mol) against known anticancer drug targets.

No.

PubChemID 1a25 1a35 1agw 1awn 1cdk 1mnc

1. CID_65064 -65.013 -100.77 -63.582 -54.635

-99.175 -119.166

2. CID_72495 -42.841 -42.841 -42.841 -42.841 -42.841 -42.841

3. CID_10794036 -80.007 -80.007 -80.007 -80.007 -80.007 -80.007

4. CID_5288209 -66.044 -66.044 -66.044 -66.044 -66.044 -66.044

5. CID_10532024 -31.384 -31.384 -31.384 -31.384 -31.384 -31.384

6. CID_401405 -47.822 -47.822 -47.822 -47.822 -47.822 -47.822

7. CID_10428109 -32.073 -32.073 -32.073 -32.073 -32.073 -32.073

8. CID_23639625 -80.799 -80.799

-80.799 -80.799

9. CID_10337961 -32.061 -32.061 -32.061 -32.061 -32.061 -32.061

10. CID_82146 -61.877 -61.877 -61.877 -61.877 -61.877 -61.877

11. CID_21773740 -65.714 -65.714 -65.714 -65.714 -65.714 -65.714

12. CID_9960816 -63.19 -63.19 -63.19 -63.19 -63.19 -63.19

13. CID_9847391 -72.317 -72.317 -72.317 -72.317 -72.317 -72.317

Note: Few docking results with modification.

Table 15.3. The library compounds violating the Lipinski’s Rule of Five and showing poor

docking interaction energy (in Kcal/mol) against known anticancer drug targets.