key: cord-0788309-ln3e9cfc authors: Chtita, Samir; Belhassan, Assia; Bakhouch, Mohamed; Taourati, Abdelali Idrissi; Aouidate, Adnane; Belaidi, Salah; Moutaabbid, Mohammed; Belaaouad, Said; Bouachrine, Mohammed; Lakhlifi, Tahar title: QSAR study of unsymmetrical aromatic Disulfides as potent avian SARS-CoV main protease inhibitors using quantum chemical descriptors and statistical methods date: 2021-02-03 journal: Chemometr Intell Lab Syst DOI: 10.1016/j.chemolab.2021.104266 sha: c549e4cfb955aed31800dc9a7e3c9a3a0d9d2dd8 doc_id: 788309 cord_uid: ln3e9cfc In silico research was executed on forty unsymmetrical aromatic disulfide derivatives as inhibitors of the SARS Coronavirus (SARS-CoV-1). Density functional theory (DFT) calculation with B3LYP functional employing 6-311+G(d,p) basis set was used to calculate quantum chemical descriptors. Topological, physicochemical and thermodynamic parameters were calculated using ChemOffice software. The dataset was divided randomly into training and test sets consisting of 32 and 8 compounds, respectively. In attempt to explore the structural requirements for bioactives molecules with significant anti-SARS-CoV-2 activity, we have built valid and robust statistics models using QSAR approach. Hundred linear pentavariate and quadrivariate models were established by changing training set compounds and further applied in test set to calculate predicted IC(50) values of compounds. Both built models were individually validated internally as well as externally along with Y-Randomization according to the OECD principles for the validation of QSAR model and the model acceptance criteria of Golbraikh and Tropsha’s. Model 34 is chosen with higher values of R(2), R(2)(test) and Q(2)cv (R(2) = 0.838, R(2)(test)= 0.735, Q(2)(cv) = 0.757). It is very important to notice that anti-SARS-CoV main protease of these compounds appear to be mainly governed by five descriptors, i.e. highest occupied molecular orbital energy (E(HOMO)), energy of molecular orbital below HOMO energy (E(HOMO-1)), Balaban index (BI), bond length between the two sulfur atoms (S1S2) and bond length between sulfur atom and benzene ring (S2Bnz). Here the possible action mechanism of these compounds was analyzed and discussed, in particular, important structural requirements for great SARS-CoV main protease inhibitor will be by substituting disulfides with smaller size electron withdrawing groups. Based on the best proposed QSAR model, some new compounds with higher SARS-CoV inhibitors activities have been designed. Further, in silico prediction studies on ADMET pharmacokinetics properties were conducted. Since its first appearance in Southern China in November 2002, the SARS coronavirus has been recognized as a global threat [1] [2] . It's an epidemic caused by severe acute respiratory syndrome SARS-CoV-1 and affected more than 8500 cases in 32 countries [3] . Symptoms are influenza-like and include high fever, malaise, myalgia, headache, non-productive cough, diarrhea, and shivering [4] . No individual symptom or cluster of symptoms has proved to be specific for a diagnosis of SARS. Although fever is the most frequently reported symptom, it is sometimes absent on initial measurement, especially in elderly and immunosuppressed patients [5] . The SARS was successfully controlled in July 2003, however, the potential reemergence of pandemic SARS-CoV is still posing a risk. In fact, the new strain of SARS (SARS-CoV-2) is potentially more virulent than the strain of 2003 outbreak [6] . SARS-CoV encodes a main protease which plays a pivotal role in processing viral polyproteins and controlling replicas complex activity. Main protease is an enzyme indispensable for viral replication and infection processes, making it an ideal target for the design of antiviral therapies [7] . In order to understand the chemical-biological interactions and predict their activities toward SARS-CoV- 1 Branowska et al. [9] had described a series of new 1,2,4-triazine unsymmetrical disulfane analogues that were prepared and evaluated as anticancer activity compounds against MCF-7 human breast cancer cells with some of them acting as low micromolar; J. K. Vandavasi et al. [10] have developed an efficient 'one pot' method for the synthesis of unsymmetrical dithio compounds directly from corresponding thiols and thiocarboxylic acids in the presence of DDQ (2,3-Dichloro-5,6-dicyano-1,4-benzoquinone). In addition, F. Yang et al. [11] have also developed one-pot synthesis of aromatic-aromatic and aromatic-aliphatic disulfide unsymmetrical disulfide using TCCA (Trichloroisocyanuric Acid). N. Stellenboom et al. [12] [13] prepared unsymmetrical glycosyl disulfides derived from sugar, alkyl/aryl or thiols. M. Bao et al. [14] have developed the N-Trifluoroacetyl arenesulfenamides effective precursors for the synthesis of unsymmetrical disulfides. Disulfides exist in many synthetic and natural products and have been applied extensively in organic transformation and medicinal chemistry. As example, ajoene and dysoxysulfone are found in garlic, onions and mahogany trees and have shown promising antifungal [15] [16] , antibacterial [17] , antitumor [9, 18] , antimalarial [19] and analgesic properties [20] . On the other hand, a literature survey reveals that several published papers describe the molecular modeling towards the main protease of SARS-CoV-1 and SARS-CoV-2 viruses. Thus, Alves et al. have performed QSAR studies to evaluate the ability of some known drugs to inhibit SARS-CoV-2 [21] . Other studies were reported by Masand et al., J o u r n a l P r e -p r o o f which describe the development of QSAR model from a dataset of peptide-type compounds as SARS-CoV inhibitors [22] [23] . The significance and novelty of findings presented in this work are reflected from the fact that we have used quantum chemistry descriptors which describe electron proprieties of the studied molecules. The use of density functional theory (DFT) is justified for the reason that some our previously QSAR studies have shown that the descriptors calculated using the DFT method can improve the accuracy of the results and lead to more reliable QSAR models [24] [25] [26] . Dataset of the inhibitor activities toward SARS Coronavirus (SARS-CoV) main protease of 40 unsymmetrical aromatic disulfides derivatives was collected from the literature [27] . Structures of the studied molecules with their activity IC 50 (μM) values are presented in Table 1 . The inhibitory activity factor IC 50 biochemical assays spectacles the required concentration of an inhibitor to achieve 50% inhibition of replication of SARS-CoV main protease. To predict the correlation between the anti-SARS-CoV activity with various quantum, topological, thermodynamic and physicochemical parameters, and to develop linear models, all the three-dimensional structures were drawn and built by GaussView 06 program [28] , quantum parameters were calculated by DFT approach performed with Gaussian 09 program package [29] using the hybrid functional B3LYP combining the Becke's three-parameter and the Lee-Yang-Parr exchange-correlation functional employing the 6-31G+(d,p) basis set in gas phase and all others parameters were calculated using Chem3D software [30] . The geometry of the compounds was determined by optimizing all geometrical variables with no symmetry constraints (Table S1 ). The pre-processing of the dataset is to eliminate the irrelevant descriptors in order to avoid the phenomenon of overfitting. Therefore, we must reduce the variables (descriptors) that do not have or have little influence on the studied activity. With the XLSTAT software [31] , we have used PCA to overview the examined compounds for similarities and dissimilarities in order to eliminate descriptors that are highly correlated and to select those that show a high correlation with the response activity; this one gives extra weight because it will be more effective at prediction. The most important result obtained by PCA is the correlation matrix, a diagonal matrix which represents the correlation between the activity and the descriptors retained. Descriptor with highest correlation is taken and compared to other descriptors in the correlation matrix. Dataset was randomly split into several training set and test set before descriptors selection. It was recommended that analysis of the models should be obtained from various splits into training set (80 %) and test set (20 %) . Then, allsubset regression for the whole dataset was obtained from the training sets and was performed using multiple linear regression (MLR) method with XLSTAT software. J o u r n a l P r e -p r o o f This method is one of the most popular methods of QSAR due to its simplicity in operation, reproducibility and ability to allow easy interpretation of the features used. The important advantage of the linear regression analysis is its transparent nature, therefore, the algorithm is accessible and predictions can be made easily [32] . Statistical parameters for modeling, internal and external validation metrics were adopted to evaluate the fit, stability and predicative power of the QSAR model. In drug discovery, the prediction of ADMET properties is an important study to escape the failure of drugs in the clinical phases [35] . Pharmacokinetic and bioavailability predictions are an essential tool in drug discovery process and should be considered to develop a new drug. Based on the pkCSM online tool [36] , the physicochemical properties of the active components were predicted, including molecular weight (MW), Partition coefficient (log P), rotatable bonds count (RB), H-bond acceptors and donors count (HBA and HBD) and polar surface area (PSA). Lipinski's rule (with MW ≤ 500 g/mol, Log P ≤ 5, NR ≤ 10, HBA ≤ 10, HBD ≤ 5, PSA ≤ 140) has been applied to evaluate the molecules drug likeness [37] . Candidate violating no more than one of these criteria is likely to be developed as a prospective oral drug [38] . Log S was also calculated to evaluate the water solubility of the proposed compounds (compound is insoluble or poorly soluble if log S ≤ -6, moderately soluble if -6 < log S ≤ -4, soluble if -4 < log S) [39] . Finally, different ADMET properties were predicted including, Absorption (Caco-2 cell permeability, P-glycoprotein (P-gp) and Human Intestinal Absorption (HIA)), Distribution (blood-brain barrier (BBB)), Metabolism (interaction of molecules with cytochrome enzyme system P450 CYP2D6 and CYP3A4), Excretion (total clearance TC)) and Toxicity (AMES toxicity, hERG I and hERG II inhibitors). These in silico pharmacokinetics parameters were evaluated to prevent the failure of those compounds during clinical studies and enhance their chances to reach the stage of being drug-candidates against the SARS-CoV-1. From the results of DFT(B3LYP/6-31G(d,p)) calculations, 11 quantum chemistry descriptors values were computed (Table S2 ). ChemOffice 3D software was used to calculate 34 others descriptors (Table S3 ). The 45 descriptors are competed for the 40 studied molecules; these descriptors were subjected to a principal component analysis. The results of this analysis are used to select the input data of multiple linear regression studies. Thus, at the beginning, we excluded all descriptors having a low correlation coefficient value (r ≤ 0.15) with respect to the dependent variable (IC 50 ). Instead, the descriptors with a correlation coefficient value greater than 0.95 are omitted to reduce the uncertainty present in our data matrix. The 25 descriptors presented in the Tables S2 and S3 are selected by the PCA analysis and used in MLR models development. QSAR analysis was performed using calculated molecular descriptors and experimental values of anti-SARS-CoV activity for the forty disulfides. Therefore, the whole dataset was randomly split into training and test sets by a good number of pentavariate and quadrivariate MLR models with nearly similar statistical performance but encompassing different descriptors (One hundred splits, 1-100) for the same size of training and test sets. Of the chemicals in the dataset, 32 compounds were selected for training set and remaining 8 compounds were considered as test set. The models that do not satisfy OECD principles [34] and Golbraikh and Tropsha's criteria [33] were summarily excluded. Fifty MLR models with highest coefficients of determination, explained variance in "leave one-out" cross validation prediction and with good ability to predict IC 50 values of test set compounds were selected for the whole dataset from all splits. The splits into training and test sets results and the performances of MLR models are shown in Tables 2 and S4 . All equations models presented in Table 2 For all developed models, values of R are quite close to R suggesting that number of descriptors in the models is not too high, thereby, indicating that the models are free from over-fitting [40] . This is further supported by the low In addition, evaluation of applicability domains of these top three models shows that only model 34 that have no responses outside or outlier in Williams plots ( Figure 1 ). Applicability domains were evaluated by leverage analysis expressed as Williams plot, in which standardized residuals and the leverage threshold values h*=0.563 (h*= 3*(k+1)/n); k=5, n=32) were plotted. Any new value of predicted pIC 50 data must be considered reliable only for J o u r n a l P r e -p r o o f those compounds that fall within this AD on which the model was constructed. Compounds with hi > h* or with standardized residual greater than y = ±3 can be considered as chemically different from the data set compounds and, thus, outside or outline the AD. From Figure 1 , it is obvious that all compounds in training and test sets satisfy outlier/outside criteria for model 34. There is no response outlier in training set and no response outside in test set; only one compound (N° 14) has a residual out of the ± 3 times standard deviation interval. In this step, all calculations were repeated with randomized activities of the training set compounds as well to evaluate model robustness (y-randomization test). In the present case, 100 random trials were run for the MLR model. None of the random trials could match the original model (Table S5 ). The standalone QSAR-tools, available online at http://teqip.jdvu.ac.in/QSAR_Tools, were employed in the y-randomization. The average value of R , R and Q ( ) are 0.413, 0.183 and -0.272 respectively, the , -value equal a 0.847, and all the new QSAR models having significantly low R and Q ( ) values for the 100 trials, which confirm that the developed QSAR models are robust. The p-value is lower than 0.0001, it means that we would be taking a lower than 0.01% risk in assuming that the null hypothesis is wrong. The high correlation coefficient R (0.915) indicates the susceptibility of descriptors (E HOMO , E HOMO-1 , BI, S2Bnz and S1S2) to form the above model and do bring a significant amount of information. Further, the generated model has achieved high activity-descriptor relationship efficiency of 84 % as shown by the regression-coefficient (R 2 =0.838). The large adjusted regression-coefficient R 2 (R 2 adj ) value presented in the generated MLR model and its closeness to the value of regression-coefficient (R 2 ) indicates that the developed model has perfect descriptive ability to descriptors in it and it further illustrates the true impact of used descriptors on the IC 50 . Cross-validated square correlation coefficient (. /0 1 ) by LOO approach was 0.757 which showed a good internal predictive ability of the model. The low R 2 and . /0 1 values obtained for all the random models by shown in Table 2 indicate that there is no chance of correlation or structural dependency in the proposed model. The high R 2 test as shown in the developed model (R 2 test =0.735) explains that the generated model can provide a good and valid prediction for the new compounds. Consequently, we can conclude with confidence that model 34 can be considered as a perfect model with both high statistical significance and excellent predictive ability and thus, can be used as a reliable tool for discovering anti-SARS-CoV with novel disulfides. Table 3 and Figure 2 . VIF values of the five descriptors are smaller than 5.0 (4.785, 3.794, 1.217, 1.266 and 1.492 for E HOMO , E HOMO-1 , BI, S2Bnz and S1S2, respectively) indicating that there is no multicollinearity among selected descriptors and resulting model has good stability [41] . Table 4 reflected the reliability and acceptability of our proposed model. In the equation of model 34, Balaban index (BI), highest occupied molecular orbital energy (E HOMO ) and bond length between the two sulfur atoms (S1S2) promote activity, while molecular orbital energy below HOMO energy (E HOMO-1 ) and bond length between sulfur atom and the benzene ring (S2Bnz) increases activity. By interpreting the descriptors contained in QSAR model, it is possible to gain some insights into factors, which are related to anti-SARS-CoV activity. For this reason, an acceptable interpretation of the selected descriptors is provided below: -Balaban index (BI) of a molecular graph calculates the average distance sum connectivity index. It is describes very well the degree of ramification of non-cyclic molecules [42] . In the model equation, BI mean effect has a positive sign in the model and variation in BI accounts for 31 % of the variance in IC 50 , which suggests that increased activity (decreased IC 50 ) can be achieved by decreasing the ramification of molecular skeleton. -The bond length between the two sulfur atoms (S1S2) has a positive sign in the model and variation in S1S2 accounts for 21 % of the variance in IC 50 , which suggests that increased activity can be achieved by substitute the molecular skeleton with stronger electron withdrawing ability group decrease S-S bond lengths. A relatively neutral or electron-withdrawing group in only one ortho position of phenyl (or any substituents at any more distant positions) allows the S-S bond to be short [43] . -The bond length between sulfur atom and benzene ring (S2Bnz) has a negative sign in the model and variation in S2Bnz accounts for 19 % of the variance in IC 50 , suggesting that increased activity can be achieved by substitute the molecular skeleton with weaker donating electron ability group that can decrease the S2Bnz bond length. The bigger the bond length between sulfur atom and benzene ring is, the weaker conjugated π system via mesomerism or inductive effects, and higher the activity is. -The energy of HOMO is directly related to the ionization potential and characterizes the susceptibility of the molecule toward attack by electrophiles. Hard nucleophiles have a low-energy HOMO, soft nucleophiles have a high energy HOMO. Hence, molecule with high energy HOMO will give up electrons more easily because it does not cost much to donate these electrons toward making a new bond [32, 44] . The contribution of E HOMO in describing anti-SARS-CoV activity may be attributed to the interaction of disulfide derivatives with nucleophilic amino acid residue of microorganisms. E HOMO has a positive sign in the model and variation in BI accounts for 19 % of the variance in IC 50 , which suggests that the higher of E HOMO , the weaker donating electron ability, is showing the fact that the nucleophilic reaction occurs more easily and the activity of the compound is higher [45] . Consequently, if we want to decrease the value of IC 50 , we will decrease E HOMO for which we must substitute the disulfide derivatives for J o u r n a l P r e -p r o o f a weaker donating electron ability group that removes electron density (don't donates density) from the conjugated π system via mesomerism effect, making it less reactive. -E HOMO-1 has a negative sign in the model. This sign suggests that the anti-SARS-CoV activity is inversely related to this descriptor. Whereas, the significance of this descriptor in the activity when its compared to the other descriptors is very weak and account for only 10 % of the variance in IC 50 . In the conclusion, these results illustrates that to increase the anti-SARS-CoV, decrease IC 50 , we will substitute the disulfide derivatives with smaller size electron withdrawing groups such as Nitro (NO 2 ), Sulfonic acid ( The proposed model using 2D-QSAR suggests that the studied activity study is highly affected by steric and electrostatic. These outcomes were supported by those obtained by L. Wang et al. [27] using CoMFA analysis. According to the above discussions, our proposed model could be applied to other unsymmetrical aromatic disulfide derivatives accordingly to Table 1 For this purpose, compounds 31 and 38 was selected as templates because they had relatively highest anti-SARS-CoV activity (IC 50 =0.516 and 0.684, respectively). The molecules were adjusted in such a way that their synthesis was experimentally achievable. Next, in-silico screen was employed by replacing various groups in R1 to R4 sites of the benzene ring; which lead to compounds with improved predicted anti-SARS-CoV activity values as shown in Table 5 . From the predicted activities, it has been observed that all the designed compounds (X1 to X12, and Y1 to Y10) have good IC 50 values compared to the 40 studied compounds in Table 1 . Compounds X3 and X6 are defined as outliers and consequently they are not be considered, because they have higher leverage which is greater than h*=0.563; we suggest all other twenty compounds for a drug-likeness and an ADMET studies. The blood-brain barrier (BBB) permeation is a prominent property in the pharmaceutical field, it helps to determine whether or not a compound can or not cross the BBB and thus exert its therapeutic effect on the brain [47] . Based on BBB report, it is clear that all proposed compounds, except X1, are capable of crossing the BBB through by passive diffusion, without upsetting the normal central nervous system (CNS) functions. P-glycoprotein (P-gp) is a trans-membrane efflux pump that transport drugs away from the cytoplasm and cell membrane causing compounds to undergo farther metabolism and clearance, thereby limiting cellular uptake of drugs resulting in therapeutic failure because the drug concentration would be lower than expected [46, 48] . The study showed that only compound Y2 can be an inhibitor for P-glycoprotein, responsible for drug effluxes and various compounds to undergo further metabolism and clearance. The intestine is normally the primary site of a drug being absorbed from an orally administered solution. This method is constructed to predict the proportion of compounds that have been absorbed through the small intestine of humans. It estimates the percentage for a given compound that will be consumed in the human intestine. A molecule with less than 30 per cent absorbance is considered poorly absorbed [48] . Based on the predicted values of HIA, all the proposed compounds can be absorbed through human intestines. The skin permeability, expressed as the skin permeability constant log (Kp), (A compound is considered to have relatively low skin permeability if it has kog Kp(cm/h)) is also an important parameter in the pharmaceutical industry to determine the risk of compounds in case there is direct contact with skin. The more negative the log (Kp) value, the less skin permeate is the molecule [49] . Hence, all proposed compounds are found to be poorly permeable to skin and accidental contact will not have any effect on the skin. Drug clearance is measured by the proportionality constant CLtot (Low value of total clearance (logCLtot) means high drug half lifetime), and occurs primarily as a combination of hepatic clearance (metabolism in the liver and biliary clearance) and renal clearance (excretion via the kidneys). It is related to bioavailability, and is important for determining dosing rates to achieve steady-state concentrations. All compounds have a low value of total clearance which means high drug half lifetime of these compounds. The Ames toxicity test is a tool commonly used to determine mutagenic ability of a compound using bacteria. A positive test indicates the compound is mutagenic, and can therefore act as a carcinogen. Most proposed new compounds, except for X1 and Y2, are likely to be AMES-negative and thus non-mutagenic. hERG of the potassium channels encoded by hERG (Human ether-a-go-go gene) are the principal causes for the development of squire long QT syndrome -leading to fatal ventricular arrhythmia. Inhibition of hERG channels has resulted in the withdrawal of many substances from the pharmaceutical market. All proposed compounds are likely to be non-hERG I/II inhibitor as shown in table 7. In conclusion, based on the Drug-likeness and ADMET studies, we suggest thirteen compounds, including X2, X3, X4, X5, X6, X7, X8, X9, X10, Y1, Y3, Y4 and Y7, which present good absorption, distribution and metabolism J o u r n a l P r e -p r o o f properties, and they present low total clearance property and show no AMES mutagenicity or hERG inhibition properties, as promising inhibitors candidates for the main protease of SARS-CoV-1 that will be synthesized and evaluated as SARS-CoV inhibitory drugs. In this study, we have used multi-MLR approaches as linear feature QSAR method to interpret the relationship between SARS-CoV inhibitory activity for forty unsymmetrical aromatic Disulfide derivatives and their chemical structural descriptors. The above QSARs study describing the anti-SARS-CoV activity of disulfides revealed that the most relevant factors to the anti-SARS-CoV activity of disulfide derivatives are steric characteristics (71% of the variance in IC 50 ) related, firstly, with the size and volume of the substituent described by Balaban index and, secondly, with the distances parameter described by the bond length between the two sulfur atoms and between sulfur atom and benzene ring, and finally by electronic characteristics (29 % of the variance in IC 50 ) related with the E HOMO and E HOMO-1 . The results suggest that derivatives of unsymmetrical aromatic Disulfide with the following structural feature may exhibit great anti-SARS-CoV activity by substituting disulfides with smaller size electron withdrawing groups. According to the developed model, the most important findings of this research are that we have designed and suggest some new compounds with possible great activities. Consequently, the proposed models can be used in anti-SARS-CoV drug research for the unsymmetrical aromatic Disulfide derivatives. ADMET evaluation shows that 13 compounds passed the stringent lead-like criteria and in silico drug-likeness test, which are excellent candidates for drug discovery and are expected to be developed as prospective oral drugs. These results encourage the collaboration with pharmacologists, academic or industrial, because the last ones many times are groping new drugs. No potential conflict of interest was reported by the authors. We would like to express our grateful to the "Agence Universitaire de la Francophone (AUF)" for funding research project (Reference: AUF-463/2020. Title: Repositionnement des médicaments et le dépistage in silico de certains composés issus des ressources naturelles pour le COVID19 via les méthodes de modélisation moléculaire). We would also like to acknowledge all our colleagues at the COVID19 Project from Morocco, Cameroun and Algeria for their amazing support, team spirit and valuable input. Design and synthesis of new tripeptide-type SARS-CoV 3CL protease inhibitors containing an electrophilic arylketone moiety Severe Acute Respiratory Syndrome) -Disease information The deadly coronaviruses: The 2003 SARS pandemic and the 2020 novel coronavirus epidemic in China Song Song and Ning Jiao, Efficient and Practical Synthesis of Unsymmetrical Disulfides via Base-Catalyzed Aerobic Oxidative Dehydrogenative Coupling of Thiols Synthesis of unsymmetrical disulfanes bearing 1,2,4-triazine scaffold and their in vitro screening towards anti-breast cancer activity Efficient synthesis of unsymmetrical disulfides Efficient one-pot construction of unsymmetrical disulfide bonds with TCCA A high-yielding, one-pot preparation of unsymmetrical glycosyl disulfides using 1-chlorobenzotriazole as an in situ trapping/oxidizing agent One-pot synthesis of unsymmetrical disulfides using 1-chlorobenzotriazole as oxidant: Interception of the sulfenyl chloride intermediate N-Trifluoroacetyl arenesulfenamides, effective precursors for synthesis of unsymmetrical disulfides and sulfonamides Disulfiram-based disulfides as narrow-spectrum antibacterial agents Extending the identification of structural features responsible for anti-SARS-CoV activity of peptide-type compounds using QSAR modeling QSAR Modeling of SARS-CoV Mpro Inhibitors Identifies Sufugolix, Cenicriviroc, Proglumetacin, and other Drugs as Candidates for Repurposing against SARS-CoV-2 Structure features of peptide-type SARS-CoV main protease inhibitors: Quantitative structure activity relationship study Quantitative structure-activity relationship studies of dibenzo[a,d]cycloalkenimine derivatives for non-competitive antagonists of N-methyl-d-aspartate based on density functional theory with electronic and topological descriptors QSPR studies of 9-aniliioacridine derivatives for their DNA drug binding properties based on density functional theory using statistical methods: Model, validation and influencing factors Mohamed Bourass, Majdouline Larif, Mohammed Bouachrine and Tahar Lakhlifi, Investigation of Antileishmanial Activities of Acridines Derivatives against Promastigotes and Amastigotes Form of Parasites Using Quantitative Structure Activity Relationship Analysis Discovery of unsymmetrical aromatic disulfides as novel inhibitors of SARS-CoV main protease: Chemical synthesis, biological evaluation, molecular docking and 3D-QSAR study QSAR study of anti-Human African Trypanosomiasis activity for 2-phenylimidazopyridines derivatives using DFT and Lipinski's descriptors, Heliyon Best practices for QSAR model development, validation, and exploitation OECD Guidance document on the validation of QSAR models Organization for Economic Co-operation & Development Predictions of the ADMET Properties of Candidate Drug Molecules Utilizing Different QSAR/QSPR Modelling Approaches pkCSM: Predicting Small-Molecule Pharmacokinetic and Toxicity Properties Using Graph-Based Signatures Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings Investigation of indirubin derivatives: a combination of 3D-QSAR, molecular docking, and ADMET towards the design of new DRAK2 inhibitors ADME Evaluation in Drug Discovery. Prediction of Aqueous Solubility Based on Atom Contribution Approach How not to develop a quantitative structure-activity or structure-property relationship (QSAR/QSPR) Multivariate Data Analysis Modélisation de molécules organiques hétérocycliques biologiquement actives par des méthodes QSAR/QSPR -Recherche de nouveaux medicaments Empirical Relations between Disulfide Bond Lengths, (N or C)-C-S-S Torsion Angles, and Substituents in Aromatic Disulfides. Crystal and Molecular Structure of 3,3'-Dihydroxydi-2-pyridyl Disulfide Quantitative structure-activity relationship studies of anticancer activity for Isatin (1H-indole-2,3-dione) derivatives based on density functional theory with electronic and topological descriptors Investigation of Antileishmanial Activities of Acridines Derivatives against Promastigotes and Amastigotes Form of Parasites Using Quantitative Structure Activity Relationship Analysis P-glycoprotein Inhibition for Optimal Drug Delivery why this drug transporter may be clinically important Ascher, pkCSM: predicting small-molecule pharmacokinetic properties using graphbased signatures In silico Pharmacokinetics and Molecular Docking Studies of Lead Compounds Derived from Diospyros Mespiliformis The eminent Rule of Five by Lipinski helps to evaluate the drug-likeness of a chemical compound or determine if a compound has the properties that would make it a potential orally active drug for humans [46] . As reported byLipinski, an orally active drug should not breach more than one of the following rules: hydrogen bond acceptor ≤ 10, octanol-water partition coefficient < 5, hydrogen bond donor ≤ 5, molecular weight <500Da and topological polar surface area <140. The results of the Lipinski's calculations using pkCSM online software are depicted in Table 6 .These results suggest that all proposed compounds show good result and are in agreement with this rule. Absorption, distribution, metabolism, excretion and toxicity (ADMET) properties of designed sulfide derivatives were predicted using pkCSM (Table 7 ).