Bioorganic & Medicinal Chemistry Letters 24 (2014) 773–779
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters journal homepage: www.elsevier.com/locate/bmcl
Molecular interactions between terpenoid mosquito repellents and human-secreted attractants Shengliang Liao a, Jie Song b,⇑, Zongde Wang a,⇑, , Jinzhu Chen a, Guorong Fan a, Zhanqian Song c, Shibin Shang c, Shangxing Chen a, Peng Wang a a b c
College of Forestry, Jiangxi Agricultural University, Nanchang 330045, Jiangxi, China Department of Chemistry and Biochemistry, University of Michigan-Flint, 303E Kearsley Street, Flint, MI 48502, USA Institute of Chemical Industry of Forest Products, Chinese Academy of Forestry, Nanjing 210042, Jiangsu, China
a r t i c l e
i n f o
Article history: Received 19 November 2013 Revised 20 December 2013 Accepted 24 December 2013 Available online 2 January 2014 Keywords: Terpenoid Mosquito repellent Human-secreted attractant Molecular interaction Repellency
a b s t r a c t Molecular interactions between terpenoid mosquito repellents and three typical human-secreted attractants, ammonia, 1-octen-3-ol, and formic acid were studied. Relative energies, bond distances, and bond angles of the molecular interactions were obtained at HF level to evaluate the interaction intensity and types. The effects of molecular interactions on repellency were investigated by the subsequent quantitative structure–activity relationship (QSAR) study. The results of this study suggest that attractant–repellent interaction should not be ignored and could be helpful for future research on the repelling mechanism of mosquito repellents. Ó 2014 Elsevier Ltd. All rights reserved.
Mosquito repellents are a group of compounds which act to prevent humans from mosquito biting.1 The spread of some fatal epidemic diseases,2 like malaria, which caused 1.17 million deaths in 2010, are caused by mosquito biting.3 Also, drug resistances become more common. For example, Plasmodium falciparum (Welch) has an increased resistance to anti-malarial drugs.4 These exemplify the reason why mosquitoes are causing more and more health issues. Therefore, as a way to remedy this situation, repellent is recommended as a way for personal protection, and thus, the development of powerful mosquito repellents is extremely important. Traditional mosquito repellent screening processes are expensive and time-consuming. Quantitative structure–activity relationship (QSAR) has therefore been applied to assist with this process.5 However, only a few studies have investigated the quantitative relationships between chemical structures of mosquito repellent and their repellency.6–11 Moreover, the repelling mechanism is still unclear and, to certain extents, controversial. The N,N-diethyl-3-methyl benzoyl amide (DEET), which was discovered in 1954,12 has been one of the most successful mosquito repellents used for decades.13 Numerous studies have been carried out in order to understand ⇑ Corresponding authors. Tel.: +1 8107623275; fax: +1 810 7666693 (J.S.); tel.: +86 13870686011 (Z.W.). E-mail addresses: [email protected] (J. Song), [email protected] (Z. Wang). Co-corresponding author. 0960-894X/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.bmcl.2013.12.102
how this repellent works. Some researchers have confirmed that DEET can block electrophysiological responses of olfactory sensory neurons to attractive odors. Davis et al. found that DEET inhibits lactic acid-sensitive neurons, a pair of chemoreceptor neurons in the grooved-peg (A3) on the antennae of the mosquito, Aedes aegypti.14 Recently, Ditzen et al. found that DEET strongly inhibited 1-octen-3-ol-evoked electrophysiological responses in Anopheles gambiae and Drosophila melanogaster.15 Another opinion is that the mosquito evades its host after its olfactory neuron is activated by repellent DEET. Syed et al. identified an olfactory receptor neuron (ORN) housed in a trichoid sensillum on the Culex quinquefasciatus antennae that detects DEET in a dose-dependent manner. This means that the mosquito endows with DEET-detecting ORNs, to detect and avoid DEET.16 Dogan et al.17 found that DEET acts as an attractant when a human host is absent and a repellent in the presence of the host. Such observations are confusing and difficult to explain. It may imply that attractants from human hosts may affect some properties of ‘commercial repellents’ and the repelling mechanism could be much more complicated than what one would expect. The role of attractants from human hosts and their potential effect on the repelling mechanism have been ignored in most QSAR studies done so far. It is known that, besides L-lactic acid, there are many attractant compounds from skin emanation, for example, ammonia, 1-octen3-ol, and some short-chain carboxylic acids. Ammonia, ranging from 17 lg/L to 17 mg/L, makes a significant contribution to the
774
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779
H
H
O
OH
N H
H
NH3
(3R)-1-Octen-3-ol
OH
Formic acid
α-pinene
β-pinene
Scheme 1. Structures of three attractants, a-pinene and b-pinene.
mosquito (Aedes aegypti) attraction behavior when placed together with lactic acid.18 1-octen-3-ol is believed to increase the attractiveness of L-lactic and CO2 in field studies.19 Bosch et al. found that C1–C3 and C5–C8 carboxylic acids, over a wide range of concentration, could enhance the attractiveness of lactic acid.20 Furthermore, Cork et al. also observed that formic acids elicit the largest amplitude EAG response in the electroantennography (EAG) assay on Anopheles gumbiae Giles.21 It is obvious that ammonia, 1-octen3-ol, some short-chain carboxylic acids, and etc make humans attractive to mosquitoes. Recently, it was hypothesized that DEET may suppress the release of physiologically relevant compounds, such as the above attractants, after smeared on human skin. Meaning, DEET altered the chemical profile of emanations by a ‘fixative’
effect that may also contribute to repellency.16 Unfortunately, more detail about the ‘fixative’ effect still remains mysterious. Because of the complicated roles of repellents and attractants, some preliminary studies have started to focus on what happens between the repellents and attractants and how they may affect the repellency.22–26 In this study, ammonia, 1-octen-3-ol, and formic acid were chosen as characteristic compounds from humans. To further understand these compounds, molecular interactions were investigated. Also, a group of terpenoid repellents as well as their effect on the mosquito repellency were studied. Theoretical calculations were performed to show how ammonia, 1-octen-3-ol, formic acid, and repellents interact between each other. Subsequent QSAR studies were used to elucidate how the complexes have an effect on the repellency. 22 Six-member-ring terpenoid mosquito repellent compounds were synthesized from a-pinene or b-pinene (Scheme 1). Their repellency against Aedes albopictus was tested. Structural information and repellency values of compounds were obtained from former studies10,27, shown in Table 1. Structures of three attractant molecules, twenty-two terpenoid mosquito repellents, and the attractant–repellent complexes were built and optimized using GaussView 4 and GAUSSIAN 03W software
Table 1 The interaction energy (in kJ/mol) calculated at HF level No.
Formula of structure
Log CRR
NH3-repellent
1-Octen-3-ol-repellent
Formic acid-repellent
O 1
CH3
O
1.767155866
8.3
15.8
20.7
1.803457116
15.3
16.7
21.0
1.861534411
13.0
8.6
25.5
1.954242509
10.3
10.2
23.1
1.908485019
19.9
10.5
21.0
1.857332
19.9
14.0
26.1
1.72427587
14.1
10.6
19.6
1.587710965
13.8
16.8
21.7
1.607455023
14.7
16.5
19.0
1.838849091
12.6
10.5
22.2
O O
C2H5
2
OH 3
OH
O O
4
H
OH O 5
CH3
O OH O
6
C2H5
O OH
7
OH O
8
O O
9
O
C2H5
OH 10
775
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779 Table 1 (continued) No.
Formula of structure
OMe
11
Log CRR
NH3-repellent
1-Octen-3-ol-repellent
Formic acid-repellent
1.716837723
12.1
13.4
16.4
1.793790385
13.7
12.8
17.2
1.860338007
13.5
11.8
17.7
1.748962861
10.2
14.2
24.1
1.783903579
14.3
16.0
21.5
1.808885867
15.0
14.6
18.2
1.5774918
12.7
8.3
15.5
1.72427587
14.3
18.5
19.4
1.754348336
14.4
17.1
19.3
OC2H5 12
O 13
O O
14
H
OCOCH3 15
OCOC2H5 16
O 17
18
CH2CH2OCOCH3
19
CH2CH2OCOC2H5 20
OCOCH3
1.700703717
12.9
17.3
18.5
21
OCOCH3
1.713490543
15.9
16.0
18.5
1.711807229
13.7
14.0
18.2
13.8
13.8
20.2
22
OCOCH3 Mean
package, respectively.28 Geometry optimizations of these conformations were done at the HF/6-31G(d) level. Interactions were considered between hydroxyl, carbonyl, and etheric groups in repellents and hydroxyl, carbonyl, and amine groups in other characteristic compounds. Characteristic compounds were allowed to rotate along repellents to have all possible conformations. After geometry optimizations, the most stable conformations were selected and geometrical parameters were obtained. Interaction energies between attractants and repellents are calculated by:
EðinteractionÞ ¼ EðARÞ EðAÞ EðRÞ E(interaction) represents the interaction energy, E(A) represents energy of the attractant, E(R) represents energy of the repellent compound and E(AR) represents energy of the complex. Figure 1 shows the structures of three complexes formed between three attractants and repellent (compound 1). Besides repellents and complexes (formed by repellents and characteristic compounds), specific fragments were defined to describe the area contributing to molecular interactions (see Fig. 1, the fraction of the molecule in red boxes were defined as fragments). Six types of descriptors, namely, constitutional descriptor, topological descriptors, geometrical descriptors, electrostatic descriptors, quantum-chemical descriptors, and thermodynamic molecular descriptors were calculated. More than 800 descriptors of repellents, fragments, and complexes (formed by repellents and characteristic compounds) were calculated by Ampac 8.16 and Codessa 2.7.10 software.
Heuristic method, encoded in Codessa 2.7.10 software, was employed to screen significant descriptors to build multilinear QSAR models. To avoid the ‘over-parameterization’ of the models, an increasing R2 value less than 0.02 was chosen as the breaking-point criterion.29 The robustness and predictive ability were validated by ‘leave-one out’ cross validation and internal validation, which are referred.7,29,30 The 22 compounds were divided into three groups for internal validation. Group A includes compounds 1, 4, 7, 10, 13, 16, 19, 22, group B includes 2, 5, 8, 11, 14, 17, 20, and group C includes 3, 6, 9, 12, 15, 18, 21. Internal three fold cross-validation was performed. The interaction energies as well as hydrogen bonds are listed in Tables 1–4 respectively. From the structures of three attractants, it is clear that formic acid could supply the carboxyl group to form the hydrogen bonding and ammonia and 1-octen-3-ol could donate the nitrogen and alcohol hydroxyl group. Because of the presence of the adjacent carbonyl moiety, the O–H group is more polarized than the O–H group of alcohols or the N–H group; thus, it enhances the dipole strength. As a result, the dipoles in carbolic acids allow these compounds to participate in energetically favorable hydrogen bonding (H-bonding) interactions with repellents, functioning as both an H-bond donor and acceptor. From Table 1, it confirms that the interaction energies between formic acid and repellents (about -20 kJ/mol) are generally stronger than those between ammonia and 1-octen-3-ol and repellents (about -14 kJ/ mol). It can be seen, from Table 4, that formic acids could form the shortest bond length and optimal bond angles.
776
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779
Figure 1. Complexes formed between three attractants and repellent.
QSAR models are obtained by linear regression method, using the descriptors screened out by heuristic method, while the cross validation coefficient and F value are employed to be the selection standard of model. Due to the collinearity control carried out during the heuristic screening, any two descriptors are found non-collinear (R2nc 60.5345, default value R2nc = 0.65).31 Table 5 shows descriptors obtained in the best QSAR models. Both ‘leave-one out’ cross-validation (LOOCV) and threefold cross-validation are performed to validate the optimal models. The results are listed in Table 6. Validations are carried out by dividing the parent data into three subsets A, B, and C, with 8, 7, and 7 compounds, respectively. Training sets are a combination of two subsets with the third one as the test set. The correlation equation using the same descriptors was derived from training sets and used to predict values for the corresponding test sets. Due to the smaller size of test sets, it is not surprising to observe the dispersive predicted R2. For cases of ammonia and 1-octen-3-ol, the average R2 (pred) of all test sets still gives satisfactory results compared to the average R2 (fit) of all training sets. For the case of formic acid, relatively large deviation of R2 (fit) from R2 (pred) is observed and it is believed due to three descriptors instead of four descriptors in the other two cases. Figure 2 gives the comparison between experimental values (log CRR) and predicted values (log CRR) from QSAR models listed in Table 5. All three cases demonstrate excellent linear relationship (R2 >0.93 for all three cases). Therefore, it is reasonable to believe that QSAR models obtained are statistically significant. Descriptors involved in the qualitative model obtained could supply some information to help us understand possible factors that may affect the repelling mechanisms. In the case of ammonia, the first statistically important descriptor is MO-related,32 which is related to HOMO energy, an indicator of the electrophobic effect.33 It is likely a factor that will affect the electronic property of the repellent and, therefore, alter the contribution to intermolecular forces. When ammonia is present, it may indicate the popularity to accept the lone pair on N. The second one is the electron–electron repulsion energy, a quantum-chemical descriptor. The third one is another MO-related descriptor, which is related to HOMO–LUMO energy gap and the stability of the repellent molecule. The last one is ZX Shadow/ZX Rectangle, a geometrical descriptor reflecting the size and geometrical shape of the
Table 2 Interaction distances and angles of ammonia-repellent complexes No.
Interaction type
Distance (Å)
Angle (°)
No.
Iinteraction type
Distance (Å)
Angle (°)
1
>NH O@C
NH O@C< >NH O@C< >NH O< >NH O< >N- HO-
2.496 2.418 2.411 2.435 2.692
102.1 107.2 106.9 105.3 169.5
13
>NH O< >NH O< >NH O< >NH O< >NH O@C
NH O< >NH O@C< >NH O< >NH O@C< >NH O< >NH O< –NH O@C< >NH O@C< >NH O@C< >NH O< >NH O
NH O@C< >NH O@C< >NH O@C
NH O< >NH O< >NH O@C
NH O@C< >NH O@C
NH O< >NH O
NH O@C< >NH O@C< >NH O< >NH O@C< >NH O@C
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters journal homepage: www.elsevier.com/locate/bmcl
Molecular interactions between terpenoid mosquito repellents and human-secreted attractants Shengliang Liao a, Jie Song b,⇑, Zongde Wang a,⇑, , Jinzhu Chen a, Guorong Fan a, Zhanqian Song c, Shibin Shang c, Shangxing Chen a, Peng Wang a a b c
College of Forestry, Jiangxi Agricultural University, Nanchang 330045, Jiangxi, China Department of Chemistry and Biochemistry, University of Michigan-Flint, 303E Kearsley Street, Flint, MI 48502, USA Institute of Chemical Industry of Forest Products, Chinese Academy of Forestry, Nanjing 210042, Jiangsu, China
a r t i c l e
i n f o
Article history: Received 19 November 2013 Revised 20 December 2013 Accepted 24 December 2013 Available online 2 January 2014 Keywords: Terpenoid Mosquito repellent Human-secreted attractant Molecular interaction Repellency
a b s t r a c t Molecular interactions between terpenoid mosquito repellents and three typical human-secreted attractants, ammonia, 1-octen-3-ol, and formic acid were studied. Relative energies, bond distances, and bond angles of the molecular interactions were obtained at HF level to evaluate the interaction intensity and types. The effects of molecular interactions on repellency were investigated by the subsequent quantitative structure–activity relationship (QSAR) study. The results of this study suggest that attractant–repellent interaction should not be ignored and could be helpful for future research on the repelling mechanism of mosquito repellents. Ó 2014 Elsevier Ltd. All rights reserved.
Mosquito repellents are a group of compounds which act to prevent humans from mosquito biting.1 The spread of some fatal epidemic diseases,2 like malaria, which caused 1.17 million deaths in 2010, are caused by mosquito biting.3 Also, drug resistances become more common. For example, Plasmodium falciparum (Welch) has an increased resistance to anti-malarial drugs.4 These exemplify the reason why mosquitoes are causing more and more health issues. Therefore, as a way to remedy this situation, repellent is recommended as a way for personal protection, and thus, the development of powerful mosquito repellents is extremely important. Traditional mosquito repellent screening processes are expensive and time-consuming. Quantitative structure–activity relationship (QSAR) has therefore been applied to assist with this process.5 However, only a few studies have investigated the quantitative relationships between chemical structures of mosquito repellent and their repellency.6–11 Moreover, the repelling mechanism is still unclear and, to certain extents, controversial. The N,N-diethyl-3-methyl benzoyl amide (DEET), which was discovered in 1954,12 has been one of the most successful mosquito repellents used for decades.13 Numerous studies have been carried out in order to understand ⇑ Corresponding authors. Tel.: +1 8107623275; fax: +1 810 7666693 (J.S.); tel.: +86 13870686011 (Z.W.). E-mail addresses: [email protected] (J. Song), [email protected] (Z. Wang). Co-corresponding author. 0960-894X/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.bmcl.2013.12.102
how this repellent works. Some researchers have confirmed that DEET can block electrophysiological responses of olfactory sensory neurons to attractive odors. Davis et al. found that DEET inhibits lactic acid-sensitive neurons, a pair of chemoreceptor neurons in the grooved-peg (A3) on the antennae of the mosquito, Aedes aegypti.14 Recently, Ditzen et al. found that DEET strongly inhibited 1-octen-3-ol-evoked electrophysiological responses in Anopheles gambiae and Drosophila melanogaster.15 Another opinion is that the mosquito evades its host after its olfactory neuron is activated by repellent DEET. Syed et al. identified an olfactory receptor neuron (ORN) housed in a trichoid sensillum on the Culex quinquefasciatus antennae that detects DEET in a dose-dependent manner. This means that the mosquito endows with DEET-detecting ORNs, to detect and avoid DEET.16 Dogan et al.17 found that DEET acts as an attractant when a human host is absent and a repellent in the presence of the host. Such observations are confusing and difficult to explain. It may imply that attractants from human hosts may affect some properties of ‘commercial repellents’ and the repelling mechanism could be much more complicated than what one would expect. The role of attractants from human hosts and their potential effect on the repelling mechanism have been ignored in most QSAR studies done so far. It is known that, besides L-lactic acid, there are many attractant compounds from skin emanation, for example, ammonia, 1-octen3-ol, and some short-chain carboxylic acids. Ammonia, ranging from 17 lg/L to 17 mg/L, makes a significant contribution to the
774
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779
H
H
O
OH
N H
H
NH3
(3R)-1-Octen-3-ol
OH
Formic acid
α-pinene
β-pinene
Scheme 1. Structures of three attractants, a-pinene and b-pinene.
mosquito (Aedes aegypti) attraction behavior when placed together with lactic acid.18 1-octen-3-ol is believed to increase the attractiveness of L-lactic and CO2 in field studies.19 Bosch et al. found that C1–C3 and C5–C8 carboxylic acids, over a wide range of concentration, could enhance the attractiveness of lactic acid.20 Furthermore, Cork et al. also observed that formic acids elicit the largest amplitude EAG response in the electroantennography (EAG) assay on Anopheles gumbiae Giles.21 It is obvious that ammonia, 1-octen3-ol, some short-chain carboxylic acids, and etc make humans attractive to mosquitoes. Recently, it was hypothesized that DEET may suppress the release of physiologically relevant compounds, such as the above attractants, after smeared on human skin. Meaning, DEET altered the chemical profile of emanations by a ‘fixative’
effect that may also contribute to repellency.16 Unfortunately, more detail about the ‘fixative’ effect still remains mysterious. Because of the complicated roles of repellents and attractants, some preliminary studies have started to focus on what happens between the repellents and attractants and how they may affect the repellency.22–26 In this study, ammonia, 1-octen-3-ol, and formic acid were chosen as characteristic compounds from humans. To further understand these compounds, molecular interactions were investigated. Also, a group of terpenoid repellents as well as their effect on the mosquito repellency were studied. Theoretical calculations were performed to show how ammonia, 1-octen-3-ol, formic acid, and repellents interact between each other. Subsequent QSAR studies were used to elucidate how the complexes have an effect on the repellency. 22 Six-member-ring terpenoid mosquito repellent compounds were synthesized from a-pinene or b-pinene (Scheme 1). Their repellency against Aedes albopictus was tested. Structural information and repellency values of compounds were obtained from former studies10,27, shown in Table 1. Structures of three attractant molecules, twenty-two terpenoid mosquito repellents, and the attractant–repellent complexes were built and optimized using GaussView 4 and GAUSSIAN 03W software
Table 1 The interaction energy (in kJ/mol) calculated at HF level No.
Formula of structure
Log CRR
NH3-repellent
1-Octen-3-ol-repellent
Formic acid-repellent
O 1
CH3
O
1.767155866
8.3
15.8
20.7
1.803457116
15.3
16.7
21.0
1.861534411
13.0
8.6
25.5
1.954242509
10.3
10.2
23.1
1.908485019
19.9
10.5
21.0
1.857332
19.9
14.0
26.1
1.72427587
14.1
10.6
19.6
1.587710965
13.8
16.8
21.7
1.607455023
14.7
16.5
19.0
1.838849091
12.6
10.5
22.2
O O
C2H5
2
OH 3
OH
O O
4
H
OH O 5
CH3
O OH O
6
C2H5
O OH
7
OH O
8
O O
9
O
C2H5
OH 10
775
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779 Table 1 (continued) No.
Formula of structure
OMe
11
Log CRR
NH3-repellent
1-Octen-3-ol-repellent
Formic acid-repellent
1.716837723
12.1
13.4
16.4
1.793790385
13.7
12.8
17.2
1.860338007
13.5
11.8
17.7
1.748962861
10.2
14.2
24.1
1.783903579
14.3
16.0
21.5
1.808885867
15.0
14.6
18.2
1.5774918
12.7
8.3
15.5
1.72427587
14.3
18.5
19.4
1.754348336
14.4
17.1
19.3
OC2H5 12
O 13
O O
14
H
OCOCH3 15
OCOC2H5 16
O 17
18
CH2CH2OCOCH3
19
CH2CH2OCOC2H5 20
OCOCH3
1.700703717
12.9
17.3
18.5
21
OCOCH3
1.713490543
15.9
16.0
18.5
1.711807229
13.7
14.0
18.2
13.8
13.8
20.2
22
OCOCH3 Mean
package, respectively.28 Geometry optimizations of these conformations were done at the HF/6-31G(d) level. Interactions were considered between hydroxyl, carbonyl, and etheric groups in repellents and hydroxyl, carbonyl, and amine groups in other characteristic compounds. Characteristic compounds were allowed to rotate along repellents to have all possible conformations. After geometry optimizations, the most stable conformations were selected and geometrical parameters were obtained. Interaction energies between attractants and repellents are calculated by:
EðinteractionÞ ¼ EðARÞ EðAÞ EðRÞ E(interaction) represents the interaction energy, E(A) represents energy of the attractant, E(R) represents energy of the repellent compound and E(AR) represents energy of the complex. Figure 1 shows the structures of three complexes formed between three attractants and repellent (compound 1). Besides repellents and complexes (formed by repellents and characteristic compounds), specific fragments were defined to describe the area contributing to molecular interactions (see Fig. 1, the fraction of the molecule in red boxes were defined as fragments). Six types of descriptors, namely, constitutional descriptor, topological descriptors, geometrical descriptors, electrostatic descriptors, quantum-chemical descriptors, and thermodynamic molecular descriptors were calculated. More than 800 descriptors of repellents, fragments, and complexes (formed by repellents and characteristic compounds) were calculated by Ampac 8.16 and Codessa 2.7.10 software.
Heuristic method, encoded in Codessa 2.7.10 software, was employed to screen significant descriptors to build multilinear QSAR models. To avoid the ‘over-parameterization’ of the models, an increasing R2 value less than 0.02 was chosen as the breaking-point criterion.29 The robustness and predictive ability were validated by ‘leave-one out’ cross validation and internal validation, which are referred.7,29,30 The 22 compounds were divided into three groups for internal validation. Group A includes compounds 1, 4, 7, 10, 13, 16, 19, 22, group B includes 2, 5, 8, 11, 14, 17, 20, and group C includes 3, 6, 9, 12, 15, 18, 21. Internal three fold cross-validation was performed. The interaction energies as well as hydrogen bonds are listed in Tables 1–4 respectively. From the structures of three attractants, it is clear that formic acid could supply the carboxyl group to form the hydrogen bonding and ammonia and 1-octen-3-ol could donate the nitrogen and alcohol hydroxyl group. Because of the presence of the adjacent carbonyl moiety, the O–H group is more polarized than the O–H group of alcohols or the N–H group; thus, it enhances the dipole strength. As a result, the dipoles in carbolic acids allow these compounds to participate in energetically favorable hydrogen bonding (H-bonding) interactions with repellents, functioning as both an H-bond donor and acceptor. From Table 1, it confirms that the interaction energies between formic acid and repellents (about -20 kJ/mol) are generally stronger than those between ammonia and 1-octen-3-ol and repellents (about -14 kJ/ mol). It can be seen, from Table 4, that formic acids could form the shortest bond length and optimal bond angles.
776
S. Liao et al. / Bioorg. Med. Chem. Lett. 24 (2014) 773–779
Figure 1. Complexes formed between three attractants and repellent.
QSAR models are obtained by linear regression method, using the descriptors screened out by heuristic method, while the cross validation coefficient and F value are employed to be the selection standard of model. Due to the collinearity control carried out during the heuristic screening, any two descriptors are found non-collinear (R2nc 60.5345, default value R2nc = 0.65).31 Table 5 shows descriptors obtained in the best QSAR models. Both ‘leave-one out’ cross-validation (LOOCV) and threefold cross-validation are performed to validate the optimal models. The results are listed in Table 6. Validations are carried out by dividing the parent data into three subsets A, B, and C, with 8, 7, and 7 compounds, respectively. Training sets are a combination of two subsets with the third one as the test set. The correlation equation using the same descriptors was derived from training sets and used to predict values for the corresponding test sets. Due to the smaller size of test sets, it is not surprising to observe the dispersive predicted R2. For cases of ammonia and 1-octen-3-ol, the average R2 (pred) of all test sets still gives satisfactory results compared to the average R2 (fit) of all training sets. For the case of formic acid, relatively large deviation of R2 (fit) from R2 (pred) is observed and it is believed due to three descriptors instead of four descriptors in the other two cases. Figure 2 gives the comparison between experimental values (log CRR) and predicted values (log CRR) from QSAR models listed in Table 5. All three cases demonstrate excellent linear relationship (R2 >0.93 for all three cases). Therefore, it is reasonable to believe that QSAR models obtained are statistically significant. Descriptors involved in the qualitative model obtained could supply some information to help us understand possible factors that may affect the repelling mechanisms. In the case of ammonia, the first statistically important descriptor is MO-related,32 which is related to HOMO energy, an indicator of the electrophobic effect.33 It is likely a factor that will affect the electronic property of the repellent and, therefore, alter the contribution to intermolecular forces. When ammonia is present, it may indicate the popularity to accept the lone pair on N. The second one is the electron–electron repulsion energy, a quantum-chemical descriptor. The third one is another MO-related descriptor, which is related to HOMO–LUMO energy gap and the stability of the repellent molecule. The last one is ZX Shadow/ZX Rectangle, a geometrical descriptor reflecting the size and geometrical shape of the
Table 2 Interaction distances and angles of ammonia-repellent complexes No.
Interaction type
Distance (Å)
Angle (°)
No.
Iinteraction type
Distance (Å)
Angle (°)
1
>NH O@C
NH O@C< >NH O@C< >NH O< >NH O< >N- HO-
2.496 2.418 2.411 2.435 2.692
102.1 107.2 106.9 105.3 169.5
13
>NH O< >NH O< >NH O< >NH O< >NH O@C
NH O< >NH O@C< >NH O< >NH O@C< >NH O< >NH O< –NH O@C< >NH O@C< >NH O@C< >NH O< >NH O
NH O@C< >NH O@C< >NH O@C
NH O< >NH O< >NH O@C
NH O@C< >NH O@C
NH O< >NH O
NH O@C< >NH O@C< >NH O< >NH O@C< >NH O@C
