304 Current Computer-Aided Drug Design, 2011, 7, 304-314

1573-4099/11 $58.00+.00 © 2011 Bentham Science Publishers Ltd.

Current Computational Approaches Towards the Rational Design of New Insecticidal Agents

Alejandro Speck-Planche*,1, Maria Natália Dias Soeiro Cordeiro*,1, Lisvey Guilarte-Montero2 and Reider Yera-Bueno2

1REQUIMTE/Department of Chemistry and Biochemistry, University of Porto, 4169-007 Porto, Portugal2Department of Chemistry, Faculty of Natural Sciences, University of Oriente, 90500 Santiago de Cuba, Cuba

Abstract: Pesticides are chemicals with a great impact in the economy of any country. They are employed for the eradication of pests. Insects constitute one of these pests which are extremely difficult to control. With the passage of the time, insects have become resistant to pesticides, causing huge crop losses and diseases in humans. For this reason, there is an increasing need for the design of more potent insecticides. The present review is focused on the current state of the application of computational approaches as essential tools for the design of novel insecticidal agents. Also, a model based on a substructural approach is presented as a rational, efficient and promising alternative for the discovery of new insecticides.

Keywords: Insecticides, QSAR, 3D-QSAR, docking, linear discriminant analysis, fragments.

1. INTRODUCTION

Pesticides constitute valuable chemicals or biological agents intended for preventing, destroying, repelling or mitigating any pest. In general terms, pests are divided in the following groups: insects, plant pathogens, weeds, mollusks, birds, mammals, fish, nematodes (roundworms), and microbes that destroy property, spread disease or are a vector for disease or cause a nuisance. Specially, resistance of insects to pesticides has evolved like a great problem which has been responsible for huge crop losses and diseases in humans. The phenomenon of resistance is due to Darwinian evolution at a rate that is accelerated by the intensity of selection from pesticides. Insects are especially well adapted for dispersal through flight, they have a high reproductive capacity with multiple generations per year, and they are often difficult to detect in products traded around the world, such as grain, flowers, or even used tires [1]. Resistant insect species include pests of agriculture and public health in approximately equal numbers (Table 1). It has been estimated that around 600 species of insects and other arthropods have developed resistance to insecticides [1, 2], and several of them, have led to serious difficulties to control in many areas around the world. Worldwide, even in developed countries, many of the pesticides discovered in the 1950s are still extensively used. Nowadays, insecticides are more potent in terms of the dose required (grams per hectare rather than kilograms per hectare) to control the pest (Fig. 1). One of the greatest problems is the increasing resistance to agrochemical pesticides by several species of insects which affect the crops of several plants with a notable economical impact. In

*Address correspondence to these authors at the REQUIMTE/Department of Chemistry and Biochemistry, University of Porto, 4169-007 Porto, Portugal; Tel: +34673119223; Fax: +351 220402659; E-mail: alejspivanovich@gmail.com (ASP); Fax: +351 220402659; E-mail: ncordeir@fc.up.pt (MNDSC)

this sense, and taking into consideration this fact, a new agrochemical will be developed today only if it is effective in protecting one or more of the following major world crops: corn, rice, soybeans, cotton, wheat, or oilseed rape [3]. Anyway, there is an urgent need for the introduction of more selective insecticides in all possible ways, particularly with different modes of action to combat the growing problems caused by resistant insects. Random screening has become less successful; consequently there has been more research, with greater resources being concentrated on areas of chemistry of proved biological activity. Only the use of computational approaches, including the use of computer graphics to provide a three-dimensional model of the active sites has been quite successful, and the number of new compounds coming onto the market has increased. This review is focused on the role of computational approaches toward the design of new insecticides. Also, we propose a model based on substructural approaches for the design, search and prediction of compounds with insecticidal activity, with emphasis on the agrochemical insecticides.

2. COMPUTATIONAL APPROACHES IN DRUG DESIGN

All methodologies which are used in drug design can be divided in two great groups. The first group is constituted by methodologies which are based on the knowledge of the three-dimensional structure of the biological receptor. They employ experimental methods such as X ray crystallography and NMR spectroscopy for the structural elucidation or homology modeling when the structure of the receptor is unknown. Those methodologies are supported by bioinformatics tools and have played a decisive role in order to provide a better understanding of the processes related to drug metabolism [4-12] and rational drug design [13-22]. On the other hand, the second group is focused on methods that

Rational Design of New Insecticidal Agents Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 305

Table 1. The Most Resistant Species of Insects to the Current Insecticides

Order Species Common Name

Amblyomma spp. Ticks

Boophilus spp. Cattle ticks

Panonychus citri Citrus red mite

Panonychus ulmi European red mite

Rhipicephalus spp. Ticks

Acarina

Tetranychus spp. Spider mites

Anoplura Pediculus capitis Head louse

Leptinotarsa decemlineata Colorado potato beetle

Oryzaephilus surinamensis Saw-toothed grain beetle

Sitophilus oryzae Rice weevil Coleoptera

Tribolium castaneum Red flour beetle

Aedes aegypti Yellowfever mosquito

Anopheles spp. Malaria mosquitoes

Culex quinquefasciatus Southern house mosquito

Haemotobia irritans Horn fly

Lucilia cuprina Sheep blow fly

Musca domestica House fly

Simulium damnosum Black fly

Diptera

Stomoxys calcitrans Stable fly

Aonidiella aurantii California red scale

Bemisia tabaci Sweet potato whitefly

Myzus persicae Peach-potato aphid

Nephotettix cincticeps Green leafhopper

Nilaparvata lugens Brown planthopper

Psylla pyricola Pear psylla

Homoptera

Trialeurodes vaporariorum Greenhouse whitefly

Heliothis armigera Cotton bollworm

Liriomyza spp. Serpentine leafminers

Plutella xylostella Diamondback moth

Sitotroga cerealella Angoumois grain moth

Lepidoptera

Spodoptera exigua Beet armyworm

Orthoptera Blattella germanica German cockroach

do not consider the three-dimensional structure of the biological receptor in explicit way. These methods can use large heterogeneous databases of compounds and have been based principally on Quantitative Structure Activity Relationship (QSAR) models [23-27], which have been strongly supported and combined with other computational approaches such as Complex Network theory (CNT) [28-34], Artificial Neural Networks (ANN) analysis [35-38], Artificial Intelligence (AI) and Supporting Vector Machine (SVM) [38-43]. Although the term "drug design" is commonly used, the most general term is design of bioactive

compounds. This means that all the above computational methodologies can be applied to the study of any kind of biological activity to correlate with the chemical structure. For this reason, these computational approaches can be extended to the design of pesticides, and in this specific case, the discovery of new insecticides.

2.1. QSAR Techniques for the Modeling of Insecticidal Activity

Several works have been reported in the field of the insecticides. Some of them have been focused on the study of physicochemical properties [44, 45], metabolism [46], toxicological profiles [47-54], modes of action [55, 56] and mechanism of resistance [57, 58] of different marketed insecticides. In the field of the rational design of insecticides, several chemical families of compounds with insecticidal activity, have been reported using QSAR techniques [55, 59-90], including 3D-QSAR methodologies such as Comparative Molecular Field Analysis (CoMFA) and Comparative Molecular Similarity Indices Analysis (CoMSIA). The aim of the application of all these techniques is the search of molecular patterns for the design of new compounds with insecticidal activity (Table 2). The discovery of more potent and versatile insecticides constitutes a challenge for the scientific community which investigates in the field of health and agricultural sciences. A research with impact in the design of insecticides was the use 3D-QSAR studies combined with docking techniques applied to benzaldehyde thiosemicarbazone, benzaldehyde, benzoic acid, and their derivatives as phenoloxidase inhibitors [82]. In this study, phenoloxidase (PO), which is known also as tyrosinase was selected as target for the design of insecticides. That is because PO is a key enzyme in insect development, responsible for catalyzing the hydroxylation of tyrosine into o-diphenols and the oxidation of o-diphenols into o-quinones. Inhibition of PO may provide a basis for novel environmentally friendly insecticides. Thus, in this work, the inhibitory activities and IC50 values of 57 compounds belonging to the benzaldehyde thiosemicarbazone, benzaldehyde, and benzoic acid families against phenoloxidase from Pieris rapae (Lepidoptera) larvae were determined. The most potent compound was 4-butylbenzaldehyde thiosemicarbazone (Fig. 2). For this reason, its inhibitory kinetics against PO in air-saturated solutions for the oxidation of L-3,4-dihydroxyphenylalanine (L-DOPA) was measured. The results indicated that the compound is a reversible non-competitive inhibitor. The bioactivity results were used to construct three-dimensional quantitative structure-activity relationship (3D-QSAR) models using CoMFA and CoMSIA approaches. After carrying out superimposition using common substructure-based alignment, robust and predictive 3D-QSAR models were obtained from CoMFA (q2=0.926, r2=0.986) and CoMSIA (q2=0.933, r2=0.984) with six optimum components. Also, the molecular interactions between the ligands and the target were studied using a flexible docking method (FlexX). The best scored candidates were docked flexibly, and the interaction between the representative compound 4-butylbenzaldehyde thiosemicarbazone and the active site was elucidated in detail.

306 Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 Speck-Planche et al.

N

HN

S

NH2

Fig. (2). Structure of 4-butylbenzaldehyde thiosemicarbazone.

One of the most promising works in the discovery of insecticides was the use of ANN in the family of spinosyns [83]. In this work was demonstrated that improvements in the efficacy and spectrum of the spinosyns, novel fermentation derived insecticides, have long been a goal within AgroSciences. The obtainment of fermentation products, which had specific modifications in the spinosyn

scaffold with improved activity, was a difficult process, since most modifications decreased the activity. Although a variety of approaches were investigated to identify new synthetic directions for the spinosyn chemistry including several explorations of the quantitative structure activity relationships (QSAR) of spinosyns, only, the application ANN to the spinosyn QSAR problem identified new directions for improved activity in the chemistry, which subsequent synthesis and testing confirmed. The ANN-based analogs coupled with other information on substitution effects resulting from spinosyn structure activity relationships lead to the discovery of spinetoram (Fig. 3).

abamectin

OH

OHO

O O

OO O

O

O

OO

O

OH

HN N

anabasine

O

O

O

pyrethrins limonene

O NH

O

carbaryl

SN

O NH

O

aldicarb

OHN

ON

aminocarb

N N N

amitraz

naphthalene

N NH

NN

H2N

NH2

cyromazine

O

O

hydroprene O

O

O

juvenile hormone I

N

HN

O

Otebufenozide

OH

O

HO

OH

HO

OH

ecdysterone

OO

precocene I

N

S SH2N

O

NH2

O

cartap

NN

HN NO2

NCl

imidacloprid

Cl

Cl

Cl

Cl ClDDT

ClCl

ClCl

ClCl

aldrin

PO

OO

O

N

Cl

Cl

Cl

fospirate

PSO

O

O

N

amiton

OO O

Cl

POO

Scoumaphos

OP

NOO

S N

diazinon

P SO

S

fonofos

OP

O NS

S

O

phosfolan

Cl

OO

FF

F

bifenthrin

Fig. (1). Some of the most important compounds which belong to different chemical families used as insecticides.

Rational Design of New Insecticidal Agents Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 307

Launched in late 2007, spinetoram provides both improved efficacy and an expanded spectrum while maintaining the exceptional environmental and toxicological profile already established for the spinosyn chemistry.

2.2. Structure-Based Drug Design and Insecticidal Activity

Due to the great importance and several applications in rational drug design, the methodologies of structure-based drug design have been extended to the search of new insecticidal agents (Table 3). These methodologies include homology modeling (HM) with sequence alignment (SA) [57, 78], molecular docking (MD) [78, 82, 91] and Quasar receptor-surface modeling (QRSM) [92].

An important contribution to the study of compounds with insecticidal activity was the use of homology modeling to study point mutations in the Drosophila melanogastercytochrome P450 CYP6A2. This protein was the key in the biochemical process in which the compound known as (1,1,1-trichloro-2,2-di(4-chlorophenyl)ethane (DDT), could be metabolized (Fig. 4). In this work, three point mutations R335S, L336V and V476L, which distinguished the sequence of a cytochrome P450 CYP6A2, were studied. These cytochrome mutations were assumed to be responsible for DDT resistance in the RDDT(R) strain of Drosophila melanogaster. To determine the impact of each mutation on the function of CYP6A2, the wild-type enzyme (CYP6A2wt) of Cyp6a2 was expressed in Escherichia coli as well as three variants carrying a single mutation, the double mutant CYP6A2vSV and the triple mutant CYP6A2vSVL. All

Table 2. Some of the Most Promising Works for the Discovery of Insecticidal Agents

Methodology Familyc Insect Target N(T/P) StatisticalIndices Authors Ref.

Non-linear QSAR (PR)

spinosyns and spinosoids Heliothis virescens 34/0 r2=0.816, q2=0.706, s=0.475 Sparks et al. [61]

Classical QSAR (MLR)

Monoterpenoids Musca domestica 20/0 r2=0.860, q2=0.720, s=0.110, F=32.59 Grodnitzky et al. [64]

Classical QSAR (MLR)

neonicotinoids Drosophila (nAChR) 20/0 r2=0.810, q2=0.704, s=0.814, F=15.91, SDEP=0.879 Debnath et al. [74]

3D-QSAR (CoMFA)

ecdysone agonists Sf-9 cell line 50/0 r2=0.928, q2=0.593, s=0.214, F=92.76 Nakagawa et al. [76]

3D-QSAR (CoMFA)

Benzl T, Benzl and BenzA derivatives Pieris rapae (PO) 45/12 r2=0.986, q2=0.926, F=448.56, SPRESS=0.257 Xue et al. [82]

3D-QSAR (CoMSIA)a Benzl T, Benzl and BenzA and derivatives

Pieris rapae (PO)

45/12 r2=0.984, q2=0.933, F=381.76, SPRESS=0.257 Xue et al. [82]

Non-linear QSAR (ANN)b spinosyns Heliothis virescens 20/0 r2=0.906, q2=0.889, s=4578 Sparks et al. [83]

3D-QSAR (CoMFA)

anthranilic diamides Plutella xylostella 32/6 r2=0.958, q2=0.785, s=0.290, F=190.49 Liu et al. [86]

3D-QSAR (CoMSIA)a anthranilic diamides Plutella xylostella 32/6 r2=0.981, q2=0.778, s=0.207, F=216.82 Liu et al. [86] a Only the best models are presented, N(T/P) - number of compounds in training and prediction series respectively, b The statistical indices are based in a graphic response of observed vs ANN calculated larval (neonate) tobacco budworm LC50s, c Chemical family, BenzlT -benzaldehyde thiosemicarbazone; Benzl -benzaldehyde; BenzA -benzoic acid; nAChR - nicotinic acetylcholine receptor; PO - Phenoloxidase, PR -Polynomial Regression, MLR - Multiple Linear Regression, r2 - coefficient of determination, q2 - cross-validation, SDEP - standard deviation of error of prediction, F is the F-coefficient, s - standard deviation.

O

OO

N

O

OH

HH

H

H

O

O

O

O

O

O

OO

N

O

OH

HH

H

H

O

O

O

O

O

major component(3'-ethoxy-5,6-dihydro spinosyn J)

minor component(3'-ethoxy spinosyn L)

Fig. (3). Structure of the components of spinetoram.

308 Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 Speck-Planche et al.

CYP6A2 variants were less stable than the CYP6A2wt protein. Two activities, enhanced in the RDDT(R) strain, were measured with all recombinant proteins, namely testosterone hydroxylation and DDT metabolism. Testosterone was hydroxylated at the 2-beta position with little quantitative variation among the variants. In contrast, metabolism of DDT was strongly affected by the mutations. The CYP6A2vSVL enzyme had an enhanced metabolism of DDT, producing dicofol, dichlorodiphenyldichloroethane and dichlorodiphenyl acetic acid. The apparent affinity of the enzymes CYP6A2wt and CYP6A2vSVL for DDT and testosterone was not significantly different as revealed by the type I difference spectra. Sequence alignments with CYP102A1 provided clues to the positions of the amino acids mutated in CYP6A2. These mutations were found spatially clustered in the vicinity of the distal end of helix I relative to the substrate recognition valley. Thus, this area including helix J is important for the structure and activity of CYP6A2. Furthermore, it was showed that point mutations in the cytochrome P450 can have a prominent role in insecticide resistance. In this sense, this work provides from one side, a better understanding of the mechanism of resistance of Drosophila melanogaster to DDT. On the other hand, this knowledge can constitute a starting point for the design of new insecticides through the inhibition of mutants of CYP6A2 in the insect above.

ClCl

Cl

Cl

Cl

Fig. (4). Structure of DDT.

3. DESIGNING OF NEW INSECTICIDES USING SUBSTRUCTURAL APPROACHES

Although the use of computational approaches and methodologies for the design of insecticides has played a vital role in the development of new compounds with insecticidal activity, some drawbacks still remain. The first fact is that almost all the computational approaches applied until now, have considered homogeneous series of compounds. This element reduces the ability to search in a wider structural diversity. On the other hand, the same techniques have been applied to model the insecticidal activity against only one species of insect. In an attempt to overcome this problem and taking into consideration the major economical and social impact of the researches

dedicated to prevent crop losses and diseases in humans, we develop a fragment-based approach for the design of insecticides using a heterogeneous database of compounds.

3.1. Methods

3.1.1. Atom-Centered Fragments

Atom-centered fragments (ACF) have demonstrated to be very useful descriptors, and have been employed in some QSAR studies [26, 93]. They provide important information about hydrophobic and dispersive interactions which are involved in biological processes such as transport and distribution of drugs through the membrane. Also, they give information about drug–receptor interactions [94]. The ACF descriptors can be defined as the number of specific atom types in a molecule. They are calculated from the molecular composition and atom connectivities. Each type of atom in the molecule is described in terms of its neighboring atoms. Hydrogen and halogen atoms are classified by the hybridization and oxidation states of the carbon atom to which they are attached. For hydrogen atoms, heteroatoms which are attached to a carbon in -position are further considered. Carbon atoms are classified by their hybridization state and depending on whether their neighbors are carbon or heteroatoms.

3.1.2. Functional Group Counts

These are other type of descriptors that express certain fragmental features. Functional group counts (FGC) are simple molecular descriptors defined as the number of specific functional groups in a molecule, and as the previous descriptors, they are also calculated from the molecular composition and atom connectivities [26, 93, 95]. The FGC descriptors represent many of the functional groups which are traditionally used in Organic Chemistry.

3.1.3. Spectral Moments of the Bond Adjacency Matrix

The approach that encloses the calculation of the spectral moments of the bond adjacency matrix is known as TOPS-MODE (TOPological Substructural MOlecular DEsign) approach and it has been applied for the description of some physicochemical properties of organic compounds [96-99]. On the other hand, the use of the TOPS-MODE approach has been extended to the study of biological activities [100-104], and also for the analysis of toxicological profiles in different families of compounds [105-111]. For the calculation of spectral moments, the molecular structure is codified by means of the edge adjacency matrix E (known also as the bond adjacency matrix B) [112]. The E matrix is a square

Table 3. Structure-Based Drug Design Methodologies Applied to the Discovery of New Insecticides

Organism Chemical Family Target Methodology Ref.

Drosophila melanogaster DDT CYP6A2 HM/SA [57]

Homo sapiens PhTrZ 3 HM and MD [78]

Pieris rapae Benzl T, Benzl and BenzA and derivatives PO MD [82]

Musca domestica sesquiterpenes derivatives GABA receptor QRSM [92] DDT - 1,1,1-trichloro-2,2-bis-(4'-chlorophenyl)ethane; PhTrZ - 1-phenyl-1H-1,2,3-triazoles; BenzlT -benzaldehyde thiosemicarbazone; Benzl -benzaldehyde; BenzA -benzoic acid; nAChR - nicotinic acetylcholine receptor; PO - Phenoloxidase, CYP6A2 - kind of cytochrome P450, 3 - human homo-oligomeric receptor, GABA - gamma-aminobutyric acid.

Rational Design of New Insecticidal Agents Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 309

symmetric matrix of order m (m being the number of chemical bonds in the molecular graph) whose elements ei jare equal to 1 if the bonds i and j are adjacent, that is they are incident to a common atom, or 0 otherwise. In order to codify information of heteroatoms, the TOPS-MODE approach uses weighted matrices E(wi j), instead of E .The weights wi j are chemically meaningful quantities such as bond distances, bond dipoles, bond polarizabilities or mathematical expressions involving atomic weights [100]. For this reason, the spectral moments of the bond adjacency matrix can be used as molecular fingerprints in QSAR studies [113-115]. By mathematical definition, the term spectral moment must be understood as the sum of the main diagonal elements (ei i ) of the natural powers of the weighted E matrix. Then, the spectral moments of order k ( k) are defined as:

k = Tr(Ek ) = (eii )k

i=1

s

(1)

where Tr means the trace of the matrix that is the sum of the diagonal entries (ei i )

k of the k-th power of the weighted Ematrix.

3.1.4. Selection of the Dataset: Calculation of the Descriptors and Development of the Model

The dataset was formed by 711 compounds (Suppl. Inf.1), where 361 of them were reported as insecticides [116]. The other 350 compounds are drugs which belong to different therapeutic categories [117]. The entire dataset was divided into training and prediction series. The training series contained 531 compounds: 272 active and 259 inactive, while the prediction series was formed by 180 compounds: 89 active and 91 inactive. The ACF and FGC calculated using DRAGON program (v5.3) [95], while the kdescriptors were calculated using Modeslab software (v1.5) [118]. In this case, k descriptors were weighted by the atomic weights and values of Abraham term related with the corrected hydrogen bond basicity. The smiles codes were used to calculate all descriptors discussed above. As the modeling technique, we selected the linear discriminant analysis (LDA) [119], to find a classification model (Eq. 3), which best describes the insecticidal activity (AINSECT), as a linear combination of the predictor X-variables (molecular descriptors Dk), with the coefficients ak. Such coefficients are optimized by means of LDA, specifically the LDA technique implemented in the STATISTICA software (version 6.0) [120], using only the training set compounds.

AINSECT = a0 + a1 D1 + a2 D2 + … + ak Dk (2) For the development of the model, AINSECT values of +1 and 1 were assigned to active and inactive compounds, respectively, but a posteriori probabilities are used instead to assert the model classification of compounds. In particular, when the probability of being active did not differ more than 5% from that of being inactive, the case was considered as unclassified by the model. The forward stepwise (FS) was applied as procedure to select the molecular descriptors with the highest influence on the insecticidal activity. This technique begins by including the variable which yields the best linear fit in terms of explaining the response. The next variable is included as that variable which most

significantly improves the existing model. Once this new model is determined, the variables included are tested to see if the model can be improved by dropping them from the model. If the model can be improved, the variable is removed and the stepwise procedure is repeated until no further variables are either included or removed. Moreover, the FS selection was subjected to the principle of parsimony. Thus, the classification model with high statistical significance, but having as few descriptors as possible, was chosen. The statistical quality of the model was estimated by examining several statistical indices, such as the Wilks’ lambda ( ), the squared of t h e Mahalanobis distance (D2), the Fisher ratio (F) and the corresponding p-level. The i s a multivariate measure of the group differences over several variables, and can take values from zero (perfect discrimination) to one (no discrimination). The D2 statistic is a measure of the separation between the active and inactive groups, and it shows if the model displays an appropriate discriminatory power for differentiating those groups. Additionally, other statistical indices were used to confirm the quality and the predictive power of the model. They were sensitivity (sens) the ability for classifying active cases, specificity (spec) the ability for classifying inactive cases and accuracy (acc) the overall prediction. These indices were determined according to the following equations:

sens = TPTA

· 100% (3)

spec = TNTI

·100% (4)

acc = TP + TNTA + TI

·100% (5)

where TP means the cases (compounds) classified correctly by the model as active, TA the total active compounds, TNmeans the cases classified correctly by the model as inactive and TI represents the total inactive compounds. Finally, we also evaluated the predictive ability of our final discriminant model by using an external set of compounds not used in the model setup. ROC Curve The sensitivity and the specificity can describe adequately the quality of a model. However, these two statistical indices have disadvantages. The most important one is that, they cannot provide information about how many time the probabilities indicate that a compound, observation or case will be predicted more as positive (active) than negative (inactive), and this is very important since it confirms together with the positive predictive value if a given case is active. However, that information can be provided by a Receiver–Operating Characteristic (ROC) analysis. ROC is a classic methodology from signal detection theory [121]. The ROC curve is created by plotting the true-positive rate against false-positive rate, or sensitivity against (1 specificity). The ROC curve going along the diagonal from bottom left to upper right represents pure-chance performance.

310 Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 Speck-Planche et al.

3.2. QSAR Discriminant Model

The best classification model derived from the training set, by combining the LDA and FS techniques along with the Dk structural representation, is given below together with the statistical parameters of the LDA: AINSECT = 4.469 · 10-3

1(Ato) – 0.804 1

(Ab-sumB20) + 0.173(Cp) – 1.800(Pyrr) – 1.697(Isoxz) – 0.260(C-006) + 1.122(C-041) – 0.509(O-057) + 2.782(P-117) – 0.653 (6) N = 531 = 0.497 D2 = 4.041 F(9,521) = 58.664 p<0.001 In Eq. 6, the model includes the three types of descriptors i.e. two k, three FGC and four ACF. Thus,

1(Ato) represents the spectral moment of order 1, weighted by

the atomic weights and 1(Ab-sumB20) is the spectral moment of

order 1, weighted by the Abraham term related with the corrected hydrogen bond basicity. In the case of FGC, the descriptor Cp means the number of fragments containing terminal primary C(sp3) atoms, Pyrr is related with the presence or absence of pyrrole rings, and Isoxz takes into consideration the presence or absence of isoxazole rings. Regarding ACF descriptors, C-041 represents the number of fragments containing a C(sp2) atom which is attached with two electronegative atoms (O, N, S, Se and halogens are considered to be electronegative atoms) by simple bonds and with one electronegative atom by double bond, C-006 is the number of fragments where a methylene carbon is attached to carbon atom and also to an electronegative atom, O-057means the number of fragments in which an oxygen atom defines phenol, enol or carboxyl groups, while P-117 takes into account the number of phosphate groups. Furthermore, the large sample size (N), high F index, and small p-value are indicative of the quality of the model. In addition, the small , and high D2 show that the model displays an adequate discriminatory power for differentiating active from inactive groups. The latter is also confirmed by the classification results; the model had a sensitivity of 85.66% and a specificity of 85.33% in the training series, for an accuracy of 85.50%. We have also examined all the compounds, searching for misclassified cases because they can be outliers. In this sense we checked the Mahalanobis distance of each molecule with respect the two centroids of both groups (actives and inactive). Generally, in the case of abnormal values, the compounds should be excluded from the model. Even though there were misclassified compounds, the deletion of them did not improve the model. In order to validate our model, we took into consideration the sensitivity, the specificity and the accuracy in prediction series. The values were 84.27% for sensitivity, 85.71% for specificity, and 85.00% for accuracy of 85%. The names or codes, and the probabilities related to insecticidal activity for all compounds in the database (expressed as percentages) are recorded in a supplementary material file 2 (Supp. Inf. 2). The areas under the ROC curves were 0.93 and 0.91 for the training and prediction series, respectively (Fig. 5). These areas can be interpreted in the following form: in the case of the training series, that value of area (0.93), means that a randomly selected compound or case from the active group will has a larger value of probability than a randomly selected compound or case from the inactive group, 93% of the times. A similar conclusion can be inferred from the

value of the area under the ROC curve in the prediction series. Altogether, this proves that our model is not a random classifier because the areas under the ROC curves are different and statistically significant from those obtained by random classifiers (area = 0.5).

Fig. (5). ROC curve.

3.2.1. Structural Interpretation of the Descriptors

QSAR models are usually employed for the prediction of molecules which have been synthesized and tested for the desired biological activity. However, if the model is used in order to design new molecular entities, the descriptors employed to construct it, should provide a clear interpretation in terms of physicochemical and/or structural properties. In the model represented by the Eq. 6, two descriptors are based on spectral moments. They take into consideration different physicochemical properties. The molecular accessibility (encoded by 1

(Ato)) is very important because its increment means that the molecule will have several regions which will be able to interact properly with the biological receptor, causing the enzymatic inhibition and the death of the insect. The diminution of regions in the molecules interacting with atoms as hydrogen bond acceptors (encoded by 1

(Ab-sum20)), will improve the hydrophobic properties of the molecule. This fact will provide that the molecule will pass through the different membranes inside the insect arriving to the biological receptor. The ACF and FGC descriptors have an easy interpretation because they indicate certain groups of atoms that form fragments and/or functional groups. The information provided by these descriptors will be strongly related with the reactivity (basicity, nucleophilic and electrophilic characteristics) or with some interactions such as those due to hydrophobic factors. Although it is not possible to determine exactly which kind of property has more influence in a given fragment which is encoded by ACF or FGC descriptors, the signs of the corresponding coefficients of these descriptors in the equation will provide an idea about the desirability of the different fragments i.e. if the fragment will be favorable or unfavorable for the development of the insecticidal activity. The most important advantage of this model is the possibility of computing the

Rational Design of New Insecticidal Agents Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 311

quantitative contributions to the insecticidal activity. Thus, we selected some fragments which are present in the molecules (Fig. 6) and their contributions to the activity under study were calculated (Table 4). The calculation of these fragment contributions provides useful information about the molecular patterns which can be essential for the development of insecticidal activity. Those fragments with positive contributions could be combined in different ways to design new molecules as possible insecticidal agents. At the same time we can extract information about the fragments with negative influence on the insecticidal activity. These fragments could be eliminated from the structure of the molecules which are insecticides. This fact could also help to increase the insecticidal activity. Table 4. Quantitative Contributions to the Insecticidal

Activity

ID Contrib. ID Contrib. ID Contrib. ID Contrib.

F1 0.305 F9 -0.303 F17 0.398 F25 -1.998

F2 0.501 F10 -0.249 F18 0.598 F26 -0.342

F3 0.436 F11 -0.158 F19 0.388 F27 -0.452

F4 -0.043 F12 0.015 F20 0.938 F28 0.777

F5 0.005 F13 1.929 F21 0.383 F29 0.030

F6 0.133 F14 2.878 F22 -0.626 F30 0.378

F7 -0.523 F15 0.809 F23 0.042 F31 0.623

F8 0.084 F16 0.496 F24 -0.181 F32 -0.425

CONCLUSIONS

Computational approaches have played a decisive role for the discovery of insecticidal agents. However, it is necessary to extend and to improve the existing in silicotechniques for the design of more potent, versatile and safer

insecticides. Special attention should be paid to the discovery of agrochemical insecticides, taking into consideration the severe economic consequences that represent crop losses worldwide. Our model which is based on a fragment-based approach, and with the use of a large heterogeneous database of compounds, is an attempt to overcome this problem. We consider that the future perspectives in the design of new and potent insecticides will be able to take more into consideration the following aspects: • Application of new approaches based on QSAR

models to combine strategies using graph-theoretical descriptors for the rational, fast and efficient prediction of the insecticidal activity in large databases of compounds.

• Use of innovative methodologies such as CNT which will permit not only essential insights for the design of new insecticides, but also will provide extremely important information about the mechanisms of resistance in species of insects to the current insecticides.

• Application of new computational approaches to study the toxicological profiles of the possible candidates to be used as insecticides.

ACKNOWLEDGEMENTS

The authors acknowledge the Portuguese Fundação para a Ciência e a Tecnologia (FCT) and the European Social Found for financial support (project PTDC/QUI-QUI/113687/ 2009 and grant SFRH/BPD/63666/2009).

SUPPLEMENTARY MATERIAL

Supplementary material is available on the publisher’s web site along with the published article.

O

N

O

N

O

NN

OH

O

SN

O SP

S

OO

Cl

O

O

O

Cl

Cl

ClP

OO

O

O

O O

SH2N

O N

N

NHN

N

N

HN

N

HN

ONH

N N

S

NH

NH

O

NH

NH

S

N

NN

F1 F2 F3 F4 F5 F6 F7 F8

F9 F10 F11 F12 F13 F14 F15 F16

F17 F18 F19 F20 F21 F22 F23 F24

F25 F26 F27 F28 F29 F30 F31 F32

O NH

O

O

Fig. (6). Some fragments which were found in the structures of the molecules.

312 Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 Speck-Planche et al.

REFERENCES

[1] Plimmer, J.R.; Gammon, D.W.; Ragsdale, N.N. Encyclopedia of Agrochemicals. John Wiley & Sons, Inc: Hoboken, New Jersey, 2003.

[2] Georghiou, G.P.; Taylor, C.E. In: Pesticide Resistance: Strategies and Tactics for Management, Glass, E.H., Ed. National Academy Press: Washington D.C., 1986; pp. 157-169.

[3] Waxman, M.F. Agrochemical and Pesticide Safety Handbook, Lewis Publishers: Boca Raton, Boston, London, New York, Washington D.C., 1998.

[4] Chou, K.C. Graphic rule for drug metabolism systems. Curr. Drug. Metab., 2010, 11, 369-378.

[5] Garcia, I.; Diop, Y.F.; Gomez, G. QSAR & complex network study of the HMGR inhibitors structural diversity. Curr. Drug. Metab., 2010, 11,307-314.

[6] Gonzalez-Diaz, H. Network topological indices, drug metabolism, and distribution. Curr. Drug. Metab., 2010, 11, 283-284.

[7] Gonzalez-Diaz, H.; Duardo-Sanchez, A.; Ubeira, F.M.; Prado-Prado, F.; Perez-Montoto, L.G.; Concu, R.; Podda, G.; Shen, B. Review of MARCH-INSIDE & complex networks prediction of drugs: ADMET, anti-parasite activity, metabolizing enzymes and cardiotoxicity proteome biomarkers. Curr. Drug. Metab., 2010, 11, 379-406.

[8] Khan, M.T. Predictions of the ADMET properties of candidate drug molecules utilizing different QSAR/QSPR modelling approaches. Curr. Drug. Metab., 2010, 11, 285-295.

[9] Martinez-Romero, M.; Vazquez-Naya, J.M.; Rabunal, J.R.; Pita-Fernandez, S.; Macenlle, R.; Castro-Alvarino, J.; Lopez-Roses, L.; Ulla, J.L.; Martinez-Calvo, A.V.; Vazquez, S.; Pereira, J.; Porto-Pazos, A.B.; Dorado, J.; Pazos, A.; Munteanu, C.R. Artificial intelligence techniques for colorectal cancer drug metabolism: ontology and complex network. Curr. Drug. Metab., 2010, 11, 347-368.

[10] Mrabet, Y.; Semmar, N. Mathematical methods to analysis of topology, functional variability and evolution of metabolic systems based on different decomposition concepts. Curr. Drug. Metab., 2010, 11, 315-341.

[11] Wang, J.F.; Chou, K.C. Molecular modeling of cytochrome P450 and drug metabolism. Curr. Drug. Metab., 2010, 11, 342-346.

[12] Zhong, W.Z.; Zhan, J.; Kang, P.; Yamazaki, S. Gender specific drug metabolism of PF-02341066 in rats--role of sulfoconjugation. Curr. Drug. Metab., 2010, 11, 296-306.

[13] Concu, R.; Podda, G.; Ubeira, F.M.; Gonzalez-Diaz, H. Review of QSAR models for enzyme classes of drug targets: Theoretical background and applications in parasites, hosts, and other organisms. Curr. Pharm. Des., 2010, 16, 2710-2723.

[14] Estrada, E.; Molina, E.; Nodarse, D.; Uriarte, E. Structural contributions of substrates to their binding to P-Glycoprotein. A TOPS-MODE approach. Curr. Pharm. Des., 2010, 16, 2676-2709.

[15] Garcia, I.; Fall, Y.; Gomez, G. QSAR, docking, and CoMFA studies of GSK3 inhibitors. Curr. Pharm. Des., 2010, 16, 2666-2675.

[16] Gonzalez-Diaz, H. QSAR and complex networks in pharmaceutical design, microbiology, parasitology, toxicology, cancer, and neurosciences. Curr. Pharm. Des., 2010, 16, 2598-2600.

[17] Gonzalez-Diaz, H.; Romaris, F.; Duardo-Sanchez, A.; Perez-Montoto, L.G.; Prado-Prado, F.; Patlewicz, G.; Ubeira, F.M. Predicting drugs and proteins in parasite infections with topological indices of complex networks: theoretical backgrounds, applications, and legal issues. Curr. Pharm. Des., 2010, 16, 2737-2764.

[18] Marrero-Ponce, Y.; Casanola-Martin, G.M.; Khan, M.T.; Torrens, F.; Rescigno, A.; Abad, C. Ligand-based computer-aided discovery of tyrosinase inhibitors. Applications of the TOMOCOMD-CARDD method to the elucidation of new compounds. Curr. Pharm. Des., 2010,16, 2601-2624.

[19] Munteanu, C.R.; Fernandez-Blanco, E.; Seoane, J.A.; Izquierdo-Novo, P.; Rodriguez-Fernandez, J.A.; Prieto-Gonzalez, J.M.; Rabunal, J.R.; Pazos, A. Drug discovery and design for complex diseases through QSAR computational methods. Curr. Pharm. Des., 2010, 16, 2640-2655.

[20] Roy, K.; Ghosh, G. Exploring QSARs with Extended Topochemical Atom (ETA) indices for modeling chemical and drug toxicity. Curr. Pharm. Des., 2010, 16, 2625-2639.

[21] Speck-Planche, A.; Scotti, M.T.; de Paulo-Emerenciano, V. Current pharmaceutical design of antituberculosis drugs: future perspectives. Curr. Pharm. Des., 2010, 16, 2656-2665.

[22] Vazquez-Naya, J.M.; Martinez-Romero, M.; Porto-Pazos, A.B.; Novoa, F.; Valladares-Ayerbes, M.; Pereira, J.; Munteanu, C.R.; Dorado, J.

Ontologies of drug discovery and design for neurology, cardiology and oncology. Curr. Pharm. Des., 2010, 16, 2724-2736.

[23] Duchowicz, P.R.; Fernandez, M.; Caballero, J.; Castro, E.A.; Fernandez, F.M. QSAR for non-nucleoside inhibitors of HIV-1 reverse transcriptase. Bioorg. Med. Chem., 2006, 14, 5876-5889.

[24] Estrada, E. How the parts organize in the whole? A top-down view of molecular descriptors and properties for QSAR and drug design. Mini Rev. Med. Chem., 2008, 8, 213-221.

[25] Gonzalez-Diaz, H.; Prado-Prado, F.; Ubeira, F.M. Predicting antimicrobial drugs and targets with the MARCH-INSIDE approach. Curr. Top. Med. Chem., 2008, 8, 1676-1690.

[26] Speck-Planche, A.; Scotti, M.T.; García-López, A.; Emerenciano, V.P.; Molina-Pérez, E.; Uriarte, E. Design of novel antituberculosis compounds using graph-theoretical and substructural approaches. Mol. Divers., 2009, 13, 445-458.

[27] Speck-Planche, A.; Scotti, M.T.; Emerenciano, V.P.; García-López, A.; Molina-Pérez, E.; Uriarte, E. Designing novel antitrypanosomal agents from a mixed graph-theoretical substructural approach. J. Comput. Chem., 2010, 31, 882-894.

[28] Estrada, E. Protein bipartivity and essentiality in the yeast protein-protein interaction network. J. Proteome Res., 2006, 5, 2177-2184.

[29] Estrada, E. Virtual identification of essential proteins within the protein interaction network of yeast. Proteomics, 2006, 6, 35-40.

[30] Prado-Prado, F.J.; Gonzalez-Diaz, H.; de la Vega, O.M.; Ubeira, F.M.; Chou, K.C. Unified QSAR approach to antimicrobials. Part 3: first multi-tasking QSAR model for input-coded prediction, structural back-projection, and complex networks clustering of antiprotozoal compounds. Bioorg. Med. Chem., 2008, 16, 5871-5880.

[31] Prado-Prado, F.J.; Martinez de la Vega, O.; Uriarte, E.; Ubeira, F.M.; Chou, K.C.; Gonzalez-Diaz, H. Unified QSAR approach to antimicrobials. 4. Multi-target QSAR modeling and comparative multi-distance study of the giant components of antiviral drug-drug complex networks. Bioorg. Med. Chem., 2009, 17, 569-575.

[32] Munteanu, C.R.; Vazquez, J.M.; Dorado, J.; Sierra, A.P.; Sanchez-Gonzalez, A.; Prado-Prado, F.J.; Gonzalez-Diaz, H. Complex network spectral moments for ATCUN motif DNA cleavage: first predictive study on proteins of human pathogen parasites. J. Proteome Res., 2009,8, 5219-5228.

[33] Vina, D.; Uriarte, E.; Orallo, F.; Gonzalez-Diaz, H. Alignment-free prediction of a drug-target complex network based on parameters of drug connectivity and protein sequence of receptors. Mol. Pharm., 2009, 6, 825-835.

[34] Garcia, I.; Munteanu, C.R.; Fall, Y.; Gomez, G.; Uriarte, E.; Gonzalez-Diaz, H. QSAR and complex network study of the chiral HMGR inhibitor structural diversity. Bioorg. Med. Chem., 2009, 17, 165-175.

[35] Fernandez, M.; Tundidor-Camba, A.; Caballero, J. Modeling of cyclin-dependent kinase inhibition by 1H-pyrazolo[3,4-d]pyrimidine derivatives using artificial neural network ensembles. J. Chem. Inf. Model., 2005, 45, 1884-1895.

[36] Gonzalez, M.P.; Caballero, J.; Helguera, A.M.; Garriga, M.; Gonzalez, G.; Fernandez, M. 2D autocorrelation modelling of the inhibitory activity of cytokinin-derived cyclin-dependent kinase inhibitors. Bull. Math. Biol., 2006, 68, 735-751.

[37] Caballero, J.; Fernandez, M. Linear and non-linear modeling of antifungal activity of some heterocyclic ring derivatives using multiple linear regression and Bayesian-regularized neural networks. J. Mol. Model., 2006, 12, 168-181.

[38] Caballero, J.; Fernandez, L.; Abreu, J.I.; Fernandez, M. Amino Acid Sequence Autocorrelation vectors and ensembles of Bayesian-Regularized Genetic Neural Networks for prediction of conformational stability of human lysozyme mutants. J. Chem. Inf. Model., 2006, 46,1255-1268.

[39] Caballero, J.; Fernandez, L.; Garriga, M.; Abreu, J.I.; Collina, S.; Fernandez, M. Proteometric study of ghrelin receptor function variations upon mutations using amino acid sequence autocorrelation vectors and genetic algorithm-based least square support vector machines. J. Mol. Graph. Model., 2007, 26, 166-178.

[40] Goulon, A.; Picot, T.; Duprat, A.; Dreyfus, G. Predicting activities without computing descriptors: graph machines for QSAR. SAR QSAR Environ. Res., 2007, 18, 141-153.

[41] Du, H.; Wang, J.; Hu, Z.; Yao, X.; Zhang, X. Prediction of fungicidal activities of rice blast disease based on least-squares support vector machines and project pursuit regression. J. Agric. Food Chem., 2008,56, 10785-10792.

[42] Munteanu, C.R.; Gonzalez-Diaz, H.; Magalhaes, A.L. Enzymes/non-enzymes classification model complexity based on composition,

Rational Design of New Insecticidal Agents Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 313

sequence, 3D and topological indices. J. Theor. Biol., 2008, 254, 476-482.

[43] Tetko, I.V.; Sushko, I.; Pandey, A.K.; Zhu, H.; Tropsha, A.; Papa, E.; Oberg, T.; Todeschini, R.; Fourches, D.; Varnek, A. Critical assessment of QSAR models of environmental toxicity against Tetrahymena pyriformis: focusing on applicability domain and overfitting by variable selection. J. Chem. Inf. Model., 2008, 48, 1733-1746.

[44] Keum, Y.S.; Kim, J.H.; Kim, Y.W.; Kim, K.; Li, Q.X. Photodegradation of diafenthiuron in water. Pest. Manag. Sci., 2002,58, 496-502.

[45] Aschi, M.; D'Archivio, A.A.; Maggi, M.A.; Mazzeo, P.; Ruggieri, F. Quantitative structure-retention relationships of pesticides in reversed-phase high-performance liquid chromatography. Anal. Chim. Acta, 2007, 582, 235-242.

[46] Kayser, H.; Eilinger, P. Metabolism of diafenthiuron by microsomal oxidation: procide activation and inactivation as mechanisms contributing to selectivity. Pest. Manag. Sci., 2001, 57, 975-980.

[47] Ito, N.; Hasegawa, R.; Imaida, K.; Kurata, Y.; Hagiwara, A.; Shirai, T. Effect of ingestion of 20 pesticides in combination at acceptable daily intake levels on rat liver carcinogenesis. Food Chem. Toxicol., 1995, 33,159-163.

[48] Roux, D.; Jooste, S.; Truter, E.; Kempster, P. An aquatic toxicological evaluation of fenthion in the context of finch control in South Africa. Ecotoxicol. Environ. Saf., 1995, 31, 164-172.

[49] Ito, N.; Hagiwara, A.; Tamano, S.; Futacuchi, M.; Imaida, K.; Shirai, T. Effects of pesticide mixtures at the acceptable daily intake levels on rat carcinogenesis. Food Chem. Toxicol., 1996, 34, 1091-1096.

[50] Eldred, D.V.; Jurs, P.C. Prediction of acute mammalian toxicity of organophosphorus pesticide compounds from molecular structure. SAR QSAR Environ. Res., 1999, 10, 75-99.

[51] Frampton, G.K. Long-term impacts of an organophosphate-based regime of pesticides on field and field-edge Collembola communities. Pest. Manag. Sci., 2002, 58, 991-1001.

[52] Hatzilazarou, S.P.; Charizopoulos, E.T.; Papadopoulou-Mourkidou, E.; Economou, A.S. Dissipation of three organochlorine and four pyrethroid pesticides sprayed in a greenhouse environment during hydroponic cultivation of gerbera. Pest. Manag. Sci., 2004, 60, 1197-1204.

[53] Lotti, M.; Moretto, A. Organophosphate-induced delayed polyneuropathy. Toxicol. Rev., 2005, 24, 37-49.

[54] Ashauer, R.; Boxall, A.; Brown, C. Uptake and elimination of chlorpyrifos and pentachlorophenol into the freshwater amphipod Gammarus pulex. Arch. Environ. Contam. Toxicol., 2006, 51, 542-548.

[55] Khambay, B.P.; Batty, D.; Jewess, P.J.; Bateman, G.L.; Hollomon, D.W. Mode of action and pesticidal activity of the natural product dunnione and of some analogues. Pest. Manag. Sci., 2003, 59, 174-182.

[56] Mastrantonio, G.; Mack, H.G.; Della Vedova, C.O. Interpretation of the mechanism of acetylcholinesterase inhibition ability by organophosphorus compounds through a new conformational descriptor. an experimental and theoretical study. J. Mol. Model., 2008,14, 813-821.

[57] Amichot, M.; Tares, S.; Brun-Barale, A.; Arthaud, L.; Bride, J.M.; Berge, J.B. Point mutations associated with insecticide resistance in the Drosophila cytochrome P450 Cyp6a2 enable DDT metabolism. Eur. J. Biochem., 2004, 271, 1250-1257.

[58] Hsu, J.C.; Wu, W.J.; Haymer, D.S.; Liao, H.Y.; Feng, H.T. Alterations of the acetylcholinesterase enzyme in the oriental fruit fly Bactrocera dorsalis are correlated with resistance to the organophosphate insecticide fenitrothion. Insect Biochem. Mol. Biol., 2008, 38, 146-154.

[59] Chilmonczyk, Z.; Ksycinska, H.; Polec, I. Application of chiral chromatographic parameters in quantitative structure-activity relationship analysis of homologous malathion derivatives. J. Chromatogr. B Biomed. Sci. Appl., 1998, 720, 65-69.

[60] Singh, A.K. Quantitative structure-activity relationships for phosphoramidothioate toxicity in housefly. Comp. Biochem. Physiol. C Pharmacol. Toxicol. Endocrinol., 1999, 123, 241-255.

[61] Sparks, T.C.; Crouse, G.D.; Durst, G. Natural products as insecticides: the biology, biochemistry and quantitative structure-activity relationships of spinosyns and spinosoids. Pest. Manag. Sci., 2001, 57,896-905.

[62] Spassova, D.P.; Singh, A.K. QSAR for acetylcholinesterase inhibition and toxicity of two classes of phosphoramidothioates. SAR QSAR Environ. Res., 2001, 11, 453-471.

[63] Ford, M.G.; Hoare, N.E.; Hudson, B.D.; Nevell, T.G.; Banting, L. QSAR studies of the pyrethroid insecticides. Part 3. A putative pharmacophore derived using methodology based on molecular

dynamics and hierarchical cluster analysis. J. Mol. Graph. Model., 2002, 21, 29-36.

[64] Grodnitzky, J.A.; Coats, J.R. QSAR evaluation of monoterpenoids' insecticidal activity. J. Agric. Food Chem., 2002, 50, 4576-4580.

[65] Jewess, P.J.; Chamberlain, K.; Boogaard, A.B.; Devonshire, A.L.; Khambay, B.P. Insecticidal 2-hydroxy-3-alkyl-1,4-naphthoquinones: correlation of inhibition of ubiquinol cytochrome c oxidoreductase (complex III) with insecticidal activity. Pest. Manag. Sci., 2002, 58,243-247.

[66] Kagabu, S.; Nishiwaki, H.; Sato, K.; Hibi, M.; Yamaoka, N.; Nakagawa, Y. Nicotinic acetylcholine receptor binding of imidacloprid-related diaza compounds with various ring sizes and their insecticidal activity against Musca domestica. Pest. Manag. Sci., 2002, 58, 483-490.

[67] Park, D.S.; Grodnitzky, J.A.; Coats, J.R. QSAR evaluation of cyanohydrins' fumigation toxicity to house fly (Musca domestica) and lesser grain borer (Rhyzopertha dominica). J. Agric. Food Chem., 2002,50, 5617-5620.

[68] Sultana, I.; Ikeda, I.; Ozoe, Y. Structure-activity relationships of benzylidene anabaseines in nicotinic acetylcholine receptors of cockroach nerve cords. Bioorg. Med. Chem., 2002, 10, 2963-2971.

[69] Bertok, B.; Pap, L.; Arvai, G.; Bakonyvari, I.; Kuruczne Ribai, Z. Structure-activity relationship study of alkynyl ether insecticide synergists and the development of MB-599 (verbutin). Pest. Manag. Sci., 2003, 59, 377-392.

[70] Bordas, B.; Komives, T.; Lopata, A. Ligand-based computer-aided pesticide design. A review of applications of the CoMFA and CoMSIA methodologies. Pest. Manag. Sci., 2003, 59, 393-400.

[71] Debnath, B.; Gayen, S.; Naskar, S.K.; Roy, K.; Jha, T. Quantitative structure-activity relationship study on some azidopyridinyl neonicotinoid insecticides for their selective affinity towards the drosophila nicotinic receptor over mammalian alpha4beta2 receptor using electrotopological state atom index. Drug. Des. Discov., 2003, 18,81-89.

[72] Kiriyama, K.; Nishiwaki, H.; Nakagawa, Y.; Nishimura, K. Insecticidal activity and nicotinic acetylcholine receptor binding of dinotefuran and its analogues in the housefly, Musca domestica. Pest. Manag. Sci., 2003, 59, 1093-1100.

[73] Morimoto, M.; Tanimoto, K.; Nakano, S.; Ozaki, T.; Nakano, A.; Komai, K. Insect antifeedant activity of flavones and chromones against Spodoptera litura. J. Agric. Food Chem., 2003, 51, 389-393.

[74] Debnath, B.; Gayen, S.; Basu, A.; Ghosh, B.; Srikanth, K.; Jha, T. Quantitative structure-activity relationship study using refractotopological state atom index on some neonicotinoid insecticides. Bioorg. Med. Chem., 2004, 12, 6137-6145.

[75] Cao, S.; Wei, N.; Zhao, C.; Li, L.; Huang, Q.; Qian, X. Syntheses, antifeedant activity, and QSAR analysis of new oxa(thia)diazolyl 3(2H)-pyridazinones. J. Agric. Food Chem., 2005, 53, 3120-3125.

[76] Nakagawa, Y.; Takahashi, K.; Kishikawa, H.; Ogura, T.; Minakuchi, C.; Miyagawa, H. Classical and three-dimensional QSAR for the inhibition of [3H]ponasterone A binding by diacylhydrazine-type ecdysone agonists to insect Sf-9 cells. Bioorg. Med. Chem., 2005, 13,1333-1340.

[77] Alam, M.S.; Kajiki, R.; Hanatani, H.; Kong, X.; Ozoe, F.; Matsui, Y.; Matsumura, F.; Ozoe, Y. Synthesis and structure-activity relationships of 1-phenyl-1H-1,2,3-triazoles as selective insect GABA receptor antagonists. J. Agric. Food Chem., 2006, 54, 1361-1372.

[78] Alam, M.S.; Huang, J.; Ozoe, F.; Matsumura, F.; Ozoe, Y. Synthesis, 3D-QSAR, and docking studies of 1-phenyl-1H-1,2,3-triazoles as selective antagonists for beta3 over alpha1beta2gamma2 GABA receptors. Bioorg. Med. Chem., 2007, 15, 5090-5104.

[79] Fujita, T.; Nakagawa, Y. QSAR and mode of action studies of insecticidal ecdysone agonists. SAR QSAR Environ. Res., 2007, 18, 77-88.

[80] Praba, G.O.; Velmurugan, D. Quantitative structure-activity relationship of some pesticides. Indian J. Biochem. Biophys., 2007, 44, 470-476.

[81] Tian, Z.; Shao, X.; Li, Z.; Qian, X.; Huang, Q. Synthesis, insecticidal activity, and QSAR of novel nitromethylene neonicotinoids with tetrahydropyridine fixed cis configuration and exo-ring ether modification. J. Agric. Food Chem., 2007, 55, 2288-2292.

[82] Xue, C.B.; Zhang, L.; Luo, W.C.; Xie, X.Y.; Jiang, L.; Xiao, T. 3D-QSAR and molecular docking studies of benzaldehyde thiosemicarbazone, benzaldehyde, benzoic acid, and their derivatives as phenoloxidase inhibitors. Bioorg. Med. Chem., 2007, 15, 2006-2015.

[83] Sparks, T.C.; Crouse, G.D.; Dripps, J.E.; Anzeveno, P.; Martynow, J.; Deamicis, C.V.; Gifford, J. Neural network-based QSAR and

314 Current Computer-Aided Drug Design, 2011, Vol. 7, No. 4 Speck-Planche et al.

insecticide discovery: spinetoram. J. Comput. Aided Mol. Des., 2008,22, 393-401.

[84] Cai, M.; Li, Z.; Fan, F.; Huang, Q.; Shao, X.; Song, G. Design and synthesis of novel insecticides based on the serotonergic ligand 1-[(4-aminophenyl)ethyl]-4-[3-(trifluoromethyl)phenyl]piperazine (PAPP). J. Agric. Food Chem., 2009, 58, 2624-2629.

[85] Hansch, C.; Verma, R.P. Larvicidal activities of some organotin compounds on mosquito larvae: a QSAR study. Eur. J. Med. Chem., 2009, 44, 260-273.

[86] Liu, G.Y.; Ju, X.L.; Cheng, J.; Liu, Z.Q. 3D-QSAR studies of insecticidal anthranilic diamides as ryanodine receptor activators using CoMFA, CoMSIA and DISCOtech. Chemosphere, 2009, 78, 300-306.

[87] Naik, P.K.; Singh, T.; Singh, H. Quantitative structure-activity relationship (QSAR) for insecticides: development of predictive in vivoinsecticide activity models. SAR QSAR Environ. Res., 2009, 20, 551-566.

[88] Xu, H.; Wang, J.; Sun, H.; Lv, M.; Tian, X.; Yao, X.; Zhang, X. Semisynthesis and quantitative structure-activity relationship (QSAR) study of novel aromatic esters of 4'-demethyl-4-deoxypodophyllotoxin as insecticidal agents. J. Agric. Food Chem., 2009, 57, 7919-7923.

[89] Liu, B.; Zhu, F.; Huang, Y.; Wang, Y.; Yu, F.; Fan, B.; Yao, J. Screening rules for leads of fungicides, herbicides, and insecticides. J. Agric. Food Chem., 2010, 58, 2673–2684.

[90] Wei, S.P.; Ji, Z.Q.; Zhang, H.X.; Zhang, J.W.; Wang, Y.H.; Wu, W.J. Isolation, biological evaluation and 3D-QSAR studies of insecticidal/narcotic sesquiterpene polyol esters. J. Mol. Model., 2011,17, 681-693.

[91] Holmwood, G.; Schindler, M. Protein structure based rational design of ecdysone agonists. Bioorg. Med. Chem., 2009, 17, 4064-4070.

[92] Schmidt, T.J.; Gurrath, M.; Ozoe, Y. Structure-activity relationships of seco-prezizaane and picrotoxane/picrodendrane terpenoids by Quasar receptor-surface modeling. Bioorg. Med. Chem., 2004, 12, 4159-4167.

[93] Speck-Planche, A.; Guilarte-Montero, L.; Yera-Bueno, R.; Rojas-Vargas, J.A.; Garcia-Lopez, A.; Uriarte, E.; Molina-Perez, E. Rational design of new agrochemical fungicides using substructural descriptors. Pest. Manag. Sci., 2011, 67, 438-445.

[94] Viswanadhan, V.N.; Reddy, M.R.; Bacquet, R.J.; Erion, M.D. Assessment of methods used for predicting lipophilicity: application to nucleosides and nucleoside bases. J. Comput. Chem., 1993, 14, 1019-1026

[95] Talete-srl. DRAGON for Windows (Software for Molecular Descriptor Calculations), v5.3; Milano Chemometrics and QSAR Research Group: Milano, 2005.

[96] Estrada, E.; Gutierrez, Y. Modeling chromatographic parameters by a novel graph theoretical sub-structural approach. J. Chromatogr. A, 1999, 858, 187-199.

[97] Estrada, E.; Gutierrez, Y.; Gonzalez, H. Modeling diamagnetic and magnetooptic properties of organic compounds with the TOSS-MODE approach. J. Chem. Inf. Comput. Sci., 2000, 40, 1386-1399.

[98] Estrada, E. A computer-based approach to describe the 13C NMR chemical shifts of alkanes by the generalized spectral moments of iterated line graphs. Comput. Chem., 2000, 24, 193-201.

[99] Estrada, E.; Diaz, G.A.; Delgado, E.J. Predicting infinite dilution activity coefficients of organic compounds in water by quantum-connectivity descriptors. J. Comput. Aided Mol. Des., 2006, 20, 539-548.

[100] Estrada, E.; Pena, A.; Garcia-Domenech, R. Designing sedative/hypnotic compounds from a novel substructural graph-theoretical approach. J. Comput. Aided Mol. Des., 1998, 12, 583-595.

[101] Estrada, E.; Peña, A. In silico studies for the rational discovery of anticonvulsant compounds. Bioorg. Med. Chem., 2000, 8, 2755-2770.

[102] Estrada, E.; Quincoces, J.A.; Patlewicz, G. Creating molecular diversity from antioxidants in Brazilian propolis. Combination of TOPS-MODE QSAR and virtual structure generation. Mol. Divers., 2004, 8, 21-33.

[103] Estrada, E.; Uriarte, E.; Montero, A.; Teijeira, M.; Santana, L.; De Clercq, E. A novel approach for the virtual screening and rational design of anticancer compounds. J. Med. Chem., 2000, 43, 1975-1985.

[104] Estrada, E.; Vilar, S.; Uriarte, E.; Gutierrez, Y. In silico studies toward the discovery of new anti-HIV nucleoside compounds with the use of TOPS-MODE and 2D/3D connectivity indices. 1. Pyrimidyl derivatives. J. Chem. Inf. Comput. Sci., 2002, 42, 1194-1203.

[105] Estrada, E. Structure-mutagenicity relationships in 2-furylethylene derivatives. A molecular orbital study of the role of nitro groups. Mutat. Res., 1998, 420, 67-75.

[106] Estrada, E.; Uriarte, E. Quantitative structure-toxicity relationships using TOPS-MODE. 1. Nitrobenzene toxicity to Tetrahymena pyriformis. SAR QSAR Environ. Res., 2001, 12, 309-324.

[107] Estrada, E.; Molina, E.; Uriarte, E. Quantitative structure-toxicity relationships using TOPS-MODE. 2. Neurotoxicity of a non-congeneric series of solvents. SAR QSAR Environ. Res., 2001, 12, 445-459.

[108] Estrada, E.; Uriarte, E.; Gutierrez, Y.; Gonzalez, H. Quantitative structure-toxicity relationships using TOPS-MODE. 3. Structural factors influencing the permeability of commercial solvents through living human skin. SAR QSAR Environ. Res., 2003, 14, 145-163.

[109] Helguera, A.M.; Gonzalez, M.P.; Cordeiro, M.N.D.S.; Perez, M.A. Quantitative structure carcinogenicity relationship for detecting structural alerts in nitroso-compounds. Toxicol. Appl. Pharmacol., 2007, 221, 189-202.

[110] Morales Helguera, A.; Perez Gonzalez, M.; Cordeiro, M.N.D.S.; Cabrera Perez, M.A. Quantitative structure-carcinogenicity relationship for detecting structural alerts in nitroso compounds: species, rat; sex, female; route of administration, gavage. Chem. Res. Toxicol., 2008, 21,633-642.

[111] Helguera, A.M.; Perez-Machado, G.; Cordeiro, M.N.D.S.; Combes, R.D. Quantitative structure-activity relationship modelling of the carcinogenic risk of nitroso compounds using regression analysis and the TOPS-MODE approach. SAR QSAR Environ. Res., 2010, 21, 277-304.

[112] Estrada, E. Edge adjacency relationship and a novel topological index related to molecular volume. J. Chem. Inf. Comput. Sci., 1995, 35 31–33.

[113] Estrada, E. Spectral moments of the edge adjacency matrix in molecular graphs. 1. Definition and applications for the prediction of physical properties of alkanes. J. Chem. Inf. Comput. Sci. 1996, 36, 844–849.

[114] Estrada, E. Spectral moments of the edge adjacency matrix in molecular graphs. 2. Molecules containing heteroatoms and QSAR applications. J. Chem. Inf. Comput. Sci., 1997, 37, 320–328.

[115] Estrada, E. Spectral moments of the edge adjacency matrix in molecular graphs. 3. Molecules containing cycles. J. Chem. Inf. Comput. Sci. 1998, 38, 23–27.

[116] Wood, A. Compendium of Pesticide Common Names. Classified List of Pesticides. Available at: http://www.alanwood.net/pesticides/ class_pesticides.html (Accessed September 23, 2011).

[117] O'Neill, M.J.; Heckelman, P.E.; Koch, C.B.; Roman, K.J. The Merck Index, An Encyclopedia of Chemicals, Drugs and Biologicals, 14th ed.; Whitehouse Station, NJ: Merck & Co., Inc: New Jersey, 2006.

[118] Estrada, E.; Gutiérrez, Y. MODESLAB, v1.5; Santiago de Compostela, 2002-2004.

[119] Van Waterbeemd, H. Chemometrics methods in molecular design,VCH, Weinhiem, 1995.

[120] StatSoft. STATISTICA. Data analysis software system, v6.0; Tulsa, 2001.

[121] Hanley, J.A.; McNeil, B.J. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology, 1982, 143,29-36.

Received: March 25, 2011 Revised: April 15, 2011 Accepted: September 12, 2011