QSAR study and theoretical investigation on the lethality of halogenated aliphatic hydrocarbons toward Aspergillus (A.) Nidulans

1.   Introduction

Halogenated aliphatic hydrocarbons (HAHs) are of growing concern because of their effectual toxic and carcinogenic features [1,2]. Nevertheless, HAHs are commonly utilized as industrial and home solvents, chemical synthesis intermediates, and have a range of other applications. Within the literature, HAHs were addressed with a potent versatility to react and covalently bind with DNA, thus causing genetic damage in versatile experimental organisms.

In this regard, developing new models capable of promptly screening for potential hazards of any conceivable substance in the aquatic habitat has become an essential demand. Among the developed models, there has been a continued interest in estimating the toxicity data via quantitative structure-activity/-toxicity relationships (QSAR/QSTR) ^[3,4,5,6]. Furthermore, the QSAR model attempts to link molecular structures to biological endpoints such as toxicity ^[7]. These findings are considerably distinct from the well-known in vivo data, which would be attributed to a number of toxic action mechanisms. In contrast to in vivo observations, in vitro toxicity data are easy to collect and often have a direct mechanistic significance. Consequently, the QSAR concept is one of the ligand-based drug design strategies that has effectively led to the development of new drug candidates for a wide range of diseases ^[8].

Based on the literature, the QSAR can be mechanistically meaningful by using suitable molecular descriptors to mediate toxicity ^[9]. Except for the toxic effects caused by receptor binding, toxicity might be described as an outcome of the chemical compound's capacity to reach and covalently interact at the active site ^[5]. Considering a common biophysical mechanism, variations in the chemical structure of a set of known compounds are linked to toxicity alterations. Thence, the QSAR is utilized to extrapolate into other compounds.

As a point of departure, Crebelli et al. ^[3] presented a QSAR model based on the molar refractivity (MR) and energy gap (E_gap) for the prediction of 55 haloalkenes towards Aspergillus (A.) nidulans. This regression model was characterized by R² and F values of 0.64 and 45.5, respectively, and outlined the effectual role of the electrophilicity in predicting the aneugenic potential of the aliphatic hydrocarbons. Furthermore, the toxicity of 52 halogenated hydrocarbons toward A. nidulans was well-characterized by Trohalaki et al. using descriptors, including polarizability (α), lowest unoccupied molecular orbital energy (E_LUMO), and molecular volume (V) ^[4]. The most preferential model was detected with an R² value of 0.79 and an F value of 0.59, thereby utilizing the B3LYP/6-31G** and HF/6-31G** levels of theory. Afterward, the lethal effect of 55 aliphatic molecules using descriptors as the logarithm of the octanol-water partition coefficient (log P) and the E_LUMO was addressed by Cronin et al. ^[10]. Subsequently, this two-descriptors-based regression model was characterized by R² and standard error values of 0.615 and 0.413, respectively.

While QSAR models are widely used, there is a lack of predictive models which specifically target the toxicity of HAHs, and this gap necessitates further research. Therefore, the current study aims to construct QSAR models to predict A. nidulans toxicity of 55 halogenated aliphatic hydrocarbons. A possible source of outlier molecules in the developed models is proposed along with the predictions.

2.   Materials and methods

2.1.   Data set

The utilized data on A. nidulans toxicity (log1/D₃₇) of 55 halogenated aliphatic hydrocarbons were previously measured. In particular, the data set splitting was performed using random division as the standard method. Based on the literature ^[11], the rational division methods produced better statistical outcomes for the test sets compared to models that used random division; however, the predictive powers of random and rational models are similar. Therefore, the selected halogenated hydrocarbons were randomly partitioned into two sets, namely, training and validation sets, which involve 41 and 14 molecules, respectively. The latter set was herein devoted to thoroughly confirming the accuracy of the developed QSTR model.

2.2.   Quantum chemical calculations and selected descriptors

Herein, all executed quantum chemical calculations were performed using the Amsterdam density functional program package (ADF 2010.02) [12,13]. Geometry optimization was first carried out for each molecule using the quadruple-ξ Slater basis set with four polarization functions (STO-QZ4P) in conjunction with the generalized gradient approximation (G_GA) within the PW91 exchange and correlation functional [14,15]. The inner electrons within the atomic shells were treated as a frozen-core approximation to accelerate the computations. In the realm of the DFT calculations, various quantum chemical parameters were calculated, including the energies of molecular orbitals (E_HOMO and E_LUMO) along with the energy gap (E_gap = E_LUMO − E_HOMO). Afterward, HyperChem (version 8.0) was employed to compute various physical properties, including the surface area grid (S) octanol-water partition coefficient (log P), molecular volume (V), molar refractivity (MR), and molecular weight (M). Additionally, the molecular ovality (O) ^[16] was computed as a fundamental indicator to represent how the molecular shape approaches a sphere, which can be defined as follows:

2.3.   Statistical analysis

For the statistical analysis, a principal component analysis (PCA) and multiple linear regression (MLR) were adopted ^[17,18,19]. To study the correlations among the variables, the PCA was executed using XLSTAT. The PCA method was used to study the individual correlation between the log1/D₃₇ and the molecular descriptors and excluded any variables that had a trivial effect on the log1/D₃₇ value (low R value). The MLR was used to generate linear models where no descriptor was excluded, and was performed using SPSS ^[20]. Due to the vast number of molecular descriptors, a stepwise multiple linear regression approach was employed to find the pertinent parameters based on the forward-selection and the backward-elimination procedures. Statistical outliers were established as those compounds with definite standardized residuals greater than 2. The feature of the developed MLR models was judged using the correlation coefficient (R²), the standard error of the estimate (s), the adjusted R² (R_adj²), and Fisher's criterion (F).

2.4.   Electrostatic potential (ESP) analysis and quantum theory of atoms in molecules (QTAIM) Calculations

The versatility of 1-bromo-2-methylpropane (1Br), 2-bromo-2-methylpropane (2Br), and 2-chloro-2-methylpropane (2Cl) to attractively interact with α-glucan (G) was thoroughly assessed and comparatively investigated. The geometry optimization calculations were first executed for the 1Br/2Br/2Cl/G monomers and the 1Br…/2Br…/2Cl…G complexes at the M06-2X/6-31+G* level of theory ^[21]. An electrostatic potential (ESP) analysis was carried out for the optimized systems to extract the molecular electrostatic potential (MEP) maps. The electron density envelope with a value of 0.002 au was employed to generate the MEP maps, as previously reported in the literature ^[22].

Relying on the optimized complexes, the interaction (E_int) energies were computed. The counterpoise correction method was invoked to eliminate the basis set superposition error (BSSE) ^[23], which is illustrated as follows:

where ${E}_{\mathbf{1B}\mathbf{r}/\mathbf{2B}\mathbf{r}/\mathbf{2C}\mathbf{l}\cdots \mathbf{G}}$, ${E}_{\mathbf{1B}\mathbf{r}/\mathbf{2B}\mathbf{r}/\mathbf{2C}\mathbf{l}\;\mathrm{i}\mathrm{n}\; \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}\mathrm{x}}$, and ${E}_{\mathbf{G}\;\mathrm{i}\mathrm{n}\; \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}\mathrm{x}}$ refer to the energies of the complex, 1Br/2Br/2Cl, and the G structures pertinent to their coordinates in the optimized complexes.

Moreover, the quantum theory of atoms in molecules (QTAIM) calculations were carried out to deeply recognize the studied interactions by extracting the bond paths (BPs) and the bond critical points (BCPs). All executed DFT computations were performed using the Gaussian 09 package ^[24]. The QTAIM analysis was conducted by utilizing the Multiwfn 3.7 ^[25] and visualized by Visual Molecular Dynamics (VMD) ^[26] packages.

3.   Results

3.1.   QSAR for the Lethality of Halogenated Aliphatic Hydrocarbons to A. nidulans

In order to select the molecular descriptors and to build the QSAR models, the PCA and MLR methods were applied. The correlation coefficients (R) between the observed Aspergillus (A.) nidulans toxicities (log1/D₃₇) of 41 halogenated aliphatic hydrocarbons and the molecular descriptors are gathered in Table 1. The PCA, including the correlation between the A. nidulans toxicities and the molecular descriptors, is illustrated in the biplot depicted in Figure 1. The data clusters were depicted by exhibiting the scores for the first components (F1) versus the second ones (F2).

Table 1.  Correlation coefficients matrix of log 1/D₃₇ and the computed parameters.

Variables	log1/D37	EHOMO	ELUMO	Egap	S	V	log P	MR	MW	α	O
log 1/D37	1
EHOMO	0.019	1
ELUMO	−0.529	0.338	1
Egap	−0.561	−0.206	0.851	1
S	0.536	0.090	0.256	0.216	1
V	0.563	0.079	0.223	0.187	0.997	1
log P	0.247	−0.202	0.183	0.303	0.672	0.667	1
MR	0.027	0.107	0.372	0.327	0.650	0.654	0.621	1
MW	0.756	−0.138	−0.753	−0.705	0.250	0.297	−0.041	−0.008	1
α	0.694	0.013	−0.001	−0.008	0.934	0.951	0.666	0.650	0.468	1
O	0.297	0.122	0.422	0.371	0.890	0.852	0.612	0.555	− 0.049	0.715	1

---

 | Show Table

DownLoad: CSV

Figure 1.  Correlation circle of log1/D₃₇ and the computed descriptors.

DownLoad: Full-Size Img PowerPoint

According to the data displayed in Figure 1, the molecular weight (MW) and the A. nidulans toxicities (log1/D₃₇) were somewhat near to each other and were highly related (R = 0.756). This significant correlation coefficient demonstrated that the MW had a prominent influence on the A. nidulans toxicities. Furthermore, the surface area grid (S) and molecular volume (V) were superimposed on each other, thereby outlining the prominent positive correlation between these molecular descriptors with R = 0.997. Additionally, the biplot revealed that the S, V, log P, α, and O had positive correlations with log1/D₃₇ (0.70 〉 R 〉 0.25).

However, E_LUMO and E_gap had negative correlations with log1/D₃₇ and were considered as important descriptors. Notwithstanding the importance of E_HOMO and MR in various QSAR studies [27,28], the E_HOMO and MR values were herein neglected due to their trivial effect on the log1/D₃₇ value, which was attributed to their low correlation coefficients (R 〈 0.03). After excluding E_HOMO and MR, the preferred regression model was constructed using a backward method based on three descriptors (See Eq 3). Particularly, considerable correlations were found between the MW and both E_LUMO and E_gap, with R values of −0.753 and −0.705, respectively.

where n represents the number of halogenated aliphatic hydrocarbon molecules.

Toward investigating the statistical significance of the selected descriptors in the considered model, the t-test values were calculated for each descriptor. Furthermore, the t-test is defined as the regression coefficient for the relevant descriptor divided by its error. The highest t-test value identifies the most important descriptors, where a molecular descriptor is considered significant when its t-test value exceeds |2|.

In model 1 (See Eq 3), the t-test values were −2.697, −2.871, and 9.589 for E_LUMO, E_gap, and S, respectively. The obtained t-test values showed that all three descriptors played an important role in the regression models, with the surface area grid being the most relevant. It is worth mentioning that the positive values of the regression coefficient outlined that the molecular descriptors in the QSTR model provided a positive behavior towards the toxicity values. Meanwhile, negative values suggested that the log1/D₃₇ values diminished by augmenting the value of the computed descriptor. From Eq 3, the log1/D₃₇ value was disclosed to decline when the E_LUMO and E_gap values increased. Notwithstanding, the enhanced S value leads, in turn, to an increase in the value of log1/D₃₇. These results were consistent with the results obtained using the PCA. The statistical results of Eq 3 for the validation set (14 molecules) were poor and were characterized by (R² = 0.443, s = 0.2449, F = 9.5267).

As mentioned earlier, the E_HOMO and MR were found to be essential descriptors for the prediction of various biological activities. Consequently, the two aforesaid descriptors were considered, which gave rise to using all the calculated descriptors in the construction of the adopted QSAR models, bearing in mind that these two insignificant descriptors (E_HOMO and MR) could account for the residual. Consequently, the preferred model was obtained based on three descriptors, which are described as follows:

The t-test values for E_gap, MR, and α were −7.080, −7.157, and 15.137, respectively, thus indicating the significant contributions of the three considered descriptors to the log1/D₃₇ value, with a highly appreciated role for the molecular polarizability descriptor. According to Eq 4, the log1/D₃₇ decreases along with increasing the E_gap and MR. On the other hand, increasing the values of the molecular polarizability causes a boost in the value of the log1/D₃₇. These results were in good agreement with those obtained using the PCA. As an essential issue, E_gap plays decisive roles in the chemical reactivity, biological activity, hydrophobicity, and electrophilicity of chemicals pertinent to the living cell activity and the associated mechanistic interactions. In that spirit, the E_gap is considered one of the most important parameters that can inversely affect the chemical reactivity.

The physical justification for using molar refractivity and polarizability in the developed QSAR models might be understood by considering that most of the biochemical process occurs in aqueous media. Therefore, the MR and α of aqueous toxic solutions are informative descriptors, which are useful in QSTR studies. As judged by the four statistical criteria (R², R_adj², s, and F values), Eq 4 is statistically superior to Eq 3. Table 2 compares the observed log1/D₃₇ values for 41 molecules to the predictions generated by Eq 4. Furthermore, Figure 2 visually depicts the relationship between these expectations and the experimental results of log1/D₃₇.

Table 2.  Observed and predicted values of log1/D₃₇ for the training set. Predictions were made using Eq 4.

Molecule name	log1/D37 (Observed)	log1/D37 (Predicted)	MAE*	Molecule name	log1/D37 (Observed)	log1/D37 (Predicted)	MAE*
Dichloromethane	–1.97	–1.96	0.17	1,3-Dichloropropene	–0.72	–0.83	0.16
Chloroform	–1.39	–1.36	0.18	1,1,3-Trichloropropene	–0.38	–0.43	0.16
Tetrachloromethane	–0.49	–0.82	0.18	3-Chloro-2-chloromethylpropene	–0.43	–0.49	0.17
1,2-Dichloroethane	–1.71	–1.64	0.18	1-Chloro-2-methylpropene	–1.10	–0.85	0.17
1,1,2-Trichloroethane	–1.03	–1.12	0.18	Chlorodibromofluoromethane	–0.49	0.07	0.17
1,1,1,2-Tetrachloroethane	–0.45	–0.56	0.18	Bromoform	–0.72	–0.58	0.14
Pentachloroethane	–0.23	–0.11	0.18	Bromochloromethane	–1.79	–1.60	0.14
Hexachloroethane	0.10	0.27	0.18	Bromotrichloromethane	0.10	–0.05	0.14
1,1,2-Trichloroethylene	–1.05	–0.68	0.19	Bromodichloromethane	–1.03	–1.02	0.14
Tetrachloroethylene	–0.08	–0.23	0.18	Chlorodibromomethane	–0.46	–0.78	0.15
1,1-Dichloroethylene	–1.40	–1.23	0.18	1-Bromo-2-chloroethane	–1.38	–1.26	0.13
1,2-Dichloropropane	–1.19	–1.21	0.18	1-Bromobutane	–0.75	–0.95	0.14
1,3-Dichloropropane	–1.02	–1.13	0.19	2-Bromo-1-chloropropane	–0.99	–0.84	0.13
1,2,3-Trichloropropane	–0.97	–0.73	0.19	1-Bromo-2-methylpropane	–1.26	–0.93	0.12
2-Chlorobutane	–1.28	–1.38	0.19	1-Bromo-4-chlorobutane	–0.52	–0.55	0.09
1,3-Dichlorobutane	–0.94	–0.76	0.19	1-Bromo-3-methylbutane	–0.49	–0.54	0.10
1-Chloro-2-methylpropane	–1.16	–1.32	0.19	Dibromodichloromethane	0.70	0.61	0.12
1-Chloropentane	–0.70	–0.95	0.19	Dibromomethane	–1.33	–1.39	0.13
1-Chlorohexane	–0.18	–0.56	0.19	Tetrabromomethane	2.00	1.79	0.16
1-Chlorooctane	–0.18	0.22	0.18	1,1,2,2-Tetrabromoethane	0.40	0.29	0.11
2,3-Dichloropropene	–0.43	–0.82	0.17
* MAE stands for mean absolute error.

---

 | Show Table

DownLoad: CSV

Figure 2.  Observed values of log1/D₃₇ for 41 molecules vis predictions made by Eq 4.

DownLoad: Full-Size Img PowerPoint

For a comparison purpose between the developed model (See Eq 4) and those constructed by Cronin et al. ^[10], the QSAR model was built for the whole data set, and the model was characterized by R² = 0.897, s = 0.219, and F = 148.766. The currently developed model was superior to those developed by Cronin et al., who reported R² = 0.615, s = 0.413, and F = 44.2 (Table S1).

3.2.   Validation set

The regression coefficients for the 41 molecules were adopted to compute the log1/D₃₇ for 14 halogenated aliphatic molecules to assess the predictive performance of the three-variable QSTR model (See Eq 4). Table 3 contains the gathered data, which is graphically displayed in Figure 3.

Table 3.  Observed and predicted values of log1/D₃₇ for the validation set. Predictions were made using Eq 4.

Molecule name	log1/D37 (Observed)	log1/D37 (Predicted)	MAE*
1,1-Dichloroethane	–1.68	–1.61	0.16
1,1,1-Trichloroethane	–1.00	–0.96	0.17
1,1,2,2-Tetrachloroethane	–0.45	–0.70	0.18
1,2-Dichloroethylene	–1.48	–1.11	0.17
2,2-Dichloropropane	–1.28	–1.19	0.15
1-Chlorobutane	–1.16	–1.31	0.16
2,3-Dichlorobutane	–0.94	–0.73	0.16
2-Chloro-2-methylpropane	–1.26	–1.32	0.15
1,1-Dichloropropene	–0.82	–0.89	0.17
3-Chloro-2-methylpropene	–0.88	–0.95	0.19
2-Bromobutane	–1.06	–0.93	0.21
1-Bromo-3-chloropropane	–0.88	–0.82	0.24
2-Bromo-2-methylpropane	–1.41	–0.86	0.33
1,2-Dibromoethylene	–0.96	–0.84	0.12
* MAE stands for mean absolute error.

---

 | Show Table

DownLoad: CSV

Figure 3.  Observed values of log1/D₃₇ for validation set versus predictions made by Eq 4. The outlier is represented as a triangle.

DownLoad: Full-Size Img PowerPoint

According to Table 3, the results were labeled with an R² value of 0.575 and a standard error of 0.214. Therefore, 2-bromo-2-methylpropane (2Br) had an outlier behavior in the verification set (See Figure 3) with a standard residual of –2.167. The mathematical performance of the model was marginally improved by excluding this particular compound from the predictive equation (R² = 0.737, s = 0.168). Still, of course, there were no posterior justifications for reselections of the data set.

In accordance with the strong statistical significance, the presented QSTR models had a substantial reliability and an internal extrapolative capacity. As illustrated in Figure 3, the observed log1/D₃₇ values for the verification set were consistent with the projected values.

3.3.   Further investigations on 2-Bromo-2-Methyl Propane

Furthermore, additional calculations were performed on 2-bromo-2-methyl propane (2Br) in order to clarify the potential causes for its outlier behavior in the QSTR models. Due to the structural similarities of 1-bromo-2-methyl propane (1Br) and 2-chloro-2-methyl propane (2Cl) to 2Br, the same procedure was performed on 2Cl and 1Br. Several studies have claimed that the outlier molecules can emerge for a variety of reasons ^{[29,30,31,32,33]}. Outliers, for instance, may be caused by inaccurate molecular descriptor values, or they may suggest an error in the experimental toxicities. They could either be the result of an uncommon mechanism or a distinct binding type. First, a deep inspection of the values of the E_gap energy was performed to examine the effect of the stability of the HAHs on the value of log 1/D₃₇. A stable molecule is indicated by an alkyl halide with a high value of E_gap, while a compound with a low E_gap value has a high effectiveness. Remarkable correlations (R = –0.561) were observed between the stability of the alkyl halide and its toxicity. More stable molecules turned out to have a high log 1/D₃₇ value. These results were previously outlined using a PCA (See Figure 1). The E_gap for 2Br, 1Br, and 2Cl were 0.187, 0.195, and 0.225 au, respectively. These values indicated that 2Br had the lowest energy gap and, therefore, demonstrated the highest efficacy (i.e., lowest stability). On the other hand, 2Cl turned out to be the most stable one. The log 1/D₃₇ values of 2Br, 1Br, and 2Cl were −1.41, −1.26, and −1.26, respectively. The identical results of log1/D₃₇ for the latter two molecules may suggest errors in the experimental toxicities' values.

3.3.1.   Interactions of halide alkyls with α-glucan

The structural compositions of A. nidulans cell walls have been studied and reported in various studies ^{[34,35,36,37,38]}. The cell wall is necessary for fungi to survive in their natural habitat ^[34]. Fungi cell walls account for approximately one-quarter of the total fungal biomass ^[35], and approximately one-third of the fungal genome is engaged in cell wall formation and maintenance ^[36]. A carbohydrate investigation of the A. nidulans wall concluded that it contains roughly 40% α-glucan and β-glucan [34,37,38]. In this study, α-glucan was utilized to explore the binding interactions between the A. nidulans cell wall and the selected halide alkyls.

3.3.2.   Electrostatic potential (ESP)analysis

An electrostatic potential (ESP) analysis was performed to unveil the nucleophilic and electrophilic regions over the molecular systems. By employing the ESP analysis, the molecular electrostatic potential (MEP) maps of the optimized monomers were extracted utilizing a 0.002 au electron density envelope. Figure 4 displays the MEP maps for the optimized geometries of 1Br, 2Br, 2Cl, and G.

Figure 4.  MEP maps of 1-bromo-2-methylpropane (1Br), 2-bromo-2-methylpropane (2Br), 2-chloro-2-methylpropane (2Cl), and α-glucan (G). The color scale is extended from the red (–0.01 au) to the blue (0.01 au) scope.

DownLoad: Full-Size Img PowerPoint

As depicted in Figure 4, the molecular surfaces with relative electrophilic and nucleophilic regions over the molecular surface of the studied systems were conspicuously observed by the presence of blue-coded surfaces (i.e., positive ESP) and red-coded surfaces (i.e., negative ESP), respectively. Notably, negative ESP regions were observed over the surfaces of the Br, Cl, and O atoms of the investigated molecules. On the other hand, the positive ESP areas were found around the surfaces of the H atoms. The obtained findings paraded the supreme penchant of the 1Br, 2Br, and 2Cl molecules to attractively interact with the G molecule through hydrogen bonding interactions.

3.3.3.   Energetic manifestations

The preferability of the 1Br, 2Br, and 2Cl molecules to interact with the G molecule was herein thoroughly addressed through the 1Br…/2Br…/2Cl…G complexes. First, geometry optimization calculations were performed. Using optimized complexes, the interaction energy (E_int) values were computed. Figure S1 shows the optimized geometries of the studied complexes within all the plausible configurations and their E_int values. Relying on the negative E_int values, the most preferable configurations of the 1Br…/2Br…/2Cl…G complexes were selected and gathered in Figure 5.

Figure 5.  Optimized complexes within the most preferable configurations of the 1Br…, 2Br…, and 2Cl…G complexes. Interaction energy (E_int) values are in kcal/mol.

DownLoad: Full-Size Img PowerPoint

As shown in Figure S1, the 1Br, 2Br, and 2Cl molecules showed a superior potentiality to interact with the G molecule through the depicted configurations with respectable negative interaction energy values. Notably, the stability of the scouted complexes was generally explained as an upshot to the presence of numerous hydrogen bonds, which were found to vary in numbers and distances. It is worth noting that the preferable interactions within the studied complexes were ascribed to the 1Br…G complex, followed by 2Br…G and 2Cl…G complexes (Figure 5). Numerically, E_int of the 1Br…, 2Br…, and 2Cl…G complexes were –10.49, –10.21, and –9.99 kcal/mol, respectively.

3.3.4.   Quantum theory of atoms in molecules (QTAIM) analysis

To gain further insight into the origin of the studied interactions, a quantum theory of atoms in molecules (QTAIM) analysis was executed for the optimized 1Br…/2Br…/2Cl…G complexes within the most preferable configurations. In that spirit, the BPs and BCPs were generated and are illustrated in Figure 6.

Figure 6.  QTAIM plots of the most preferable configurations of the optimized 1Br…, 2Br…, and 2Cl…G complexes.

DownLoad: Full-Size Img PowerPoint

As shown in Figure 6, the occurrence of interactions within the investigated complexes was assured by the presence of BPs and BCPs (i.e., hydrogen bonds). The higher number of the BPs and BCPs within the 1Br…G complex over the other analogs could interpret its higher preferability in comparison to the others. This observation was in line with the interaction energy affirmations.

4.   Conclusions

Aspergillus (A.) nidulans toxicities (log1/D₃₇) were precisely predicted for a set of 55 halogenated aliphatic hydrocarbons (HAHs) using DFT calculations. Besides, the individual correlations between a series of evaluated descriptors and the A. nidulans toxicities (log1/D₃₇) were unveiled. The importance of quantum-chemical parameters along with various physical properties, including the surface area grid (S), molecular volume (V), molar refractivity (MR), molecular weight (MW), and molecular ovality (O) were investigated to predict the A. nidulans toxicities of halogenated aliphatic hydrocarbons. A potential three-variable QSTR model was trained using 41 molecules and subsequently validated for a set of 14 compounds. Based on the E_gap, MR, and α, the linear model had correlation coefficient (R²) values of 0.912 and 0.575 for the training and the validation sets, respectively. The generated QSAR model could be used for the log1/D₃₇ predictions of HAHs provided in this paper; however, more research would be required to test more aliphatic compounds for an external validation. The most preferred interactions were clearly identified within the 1-bromo-2-methylpropane…α-glucan complex. The greater number of bond pathways and bond critical sites within the 1-bromo-2-methylpropane…α-glucan complex over other analogs affirmed its significant favorability.

Use of AI tools declaration

The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

Conflict of interest

All authors have no conflict of interest to report.