Elucidation on the Physicochemical Properties of Potential and Clinically Approved Antiviral Drugs: A Search for Effective Therapies against SARS-CoV-2 Infection

COVID-19 has been confirmed in millions of individuals worldwide, rendering it a global medical emergency. In the absence of vaccines and the unavailability of effective drugs for the SARS-CoV-2 infection, vaccine development is being continuously explored and several antiviral compounds and immunotherapies are currently being investigated. Given the high similarity in genetic identity between SARS-CoV and SARS-CoV-2, the present investigation identified the interaction between the physicochemical properties and the antiviral activity of different potential and clinically approved antiviral drugs against SARS-CoV using hierarchically weighted principal component analysis. Representative drugs from the classes of neuraminidase inhibitors, reverse transcriptase inhibitors, protease inhibitors, nucleoside analogues, and other compounds with potential antiviral activity were examined. The pharmacologic classification and the biological activity of the different antiviral drugs were described using indices, namely, rotatable bond count, molecular weight, heavy atom count, and molecular complexity (92.32% contribution rate). The physicochemical properties and inhibitory action against SARS-CoV-2 of lopinavir, chloroquine, ivermectin, and ciclesonide validated the adequacy of the current computational approach. The findings of the present study provide additional information, although further investigation is warranted to identify potential targets and establish exact mechanisms, in the emergent search and design of antiviral drug candidates and their subsequent synthesis as effective therapies for COVID-19.


INTRODUCTION
SARS-CoV-2, a novel severe acute respiratory syndrome coronavirus 2, has been identified to cause coronavirus disease 2019 (COVID-19) 1 . COVID-19 is a recent medical emergency worldwide with more than 3.5 million confirmed cases and more than a quarter of a million deaths as of May 4, 2020 2 . Among infected patients, supportive care to help alleviate the symptoms has been recommended 3 . At present, neither effective drugs exist nor vaccines are available for COVID-19; however, vaccines are being developed and several antiviral agents, chemotherapeutics, and immunotherapies are being investigated as pharmacologic interventions.
Despite the very complex process, the search for effective therapies for COVID-19 continues. As SARS-CoV and SARS-CoV-2 have high similarity in genetic identity 4,5 , drugs with inhibitory action against SARS-CoV would exhibit similar degrees of inhibition against SARS-CoV-2. In the present investigation, representative drugs from the classes of neuraminidase inhibitors (NAIs), reverse transcriptase inhibitors (RTIs), protease inhibitors, nucleoside analogues, and other compounds with potential antiviral activity were examined. The chemical and physical properties of drugs such as hydrogen bond donor and acceptor counts, topological polar surface area, heavy atom and rotatable bond counts, complexity, and molecular weight were identified to influence their biological activities 6,7 . Hence, the present study identified the relationship between the physicochemical properties and the antiviral activity against SARS-CoV of the available potential and clinically approved antiviral drugs using hierarchically weighted principal component analysis. Identifying the relationship between the properties of a compound and its biological activity has always been considered important in drug design 8 . The generated relationship will identify significant chemical and physical properties of these antiviral drugs explaining inhibition variations against SARS-CoV-2. The findings of the present investigation offer additional insights relevant to the search for potential antiviral drug candidates and their subsequent synthesis as effective therapies for COVID-19.

MATERIAlS AND METHODS Chemical and physical properties of drugs with antiviral activity
Information on the potential and clinically approved antiviral drugs against SARS-CoV was retrieved from the literature. These are representative drugs from the classes of NAIs, RTIs, protease inhibitors, nucleoside analogues, and compounds with potential antiviral activity. In the class of nucleoside analogues, representative drugs such as acyclovir, foscavir, and ganciclovir were identified. Another major antiviral pharmacologic class examined was the human immunodeficiency virus (HIV) antiretroviral drugs, including indinavir, nelfinavir, and saquinavir as protease inhibitors. For the HIV antiretroviral drugs classified as RTIs, selected HIV nucleoside RTIs included in this study were lamivudine and zidovudine. The fourth pharmacologic class of the antivirals examined was the NAIs with oseltamivir and zanamivir as representative drugs. In addition to these clinically approved antiviral drugs, 10 other commercially available drugs with potential antiviral activity were also included. The drugs which exhibited in vitro antiviral activity against SARS-CoV were amantadine 9 , calpain inhibitor III 10 , calpain inhibitor VI 11 , chloroquine 12 , cinanserin 13 , glycyrrhizin 14 , mizoribine 15 , niclosamide 16 , ribavirin 9,10,15 , and valinomycin 17 . The chemical and physical properties of these compounds, namely topological polar surface area, heavy atom count, hydrogen bond acceptor count, hydrogen bond donor count, complexity, rotatable bond count, and molecular weight were retrieved from online databases 18-37 (Table 1).

Principal component analysis and computational validation
Similarity in the biological activity and pharmacologic classification of the different potential and clinically approved antiviral drugs were determined, and the presence of correlations among the various chemical and physical properties of these antiviral drugs was identified using principal component analysis. A principal component contains uncorrelated linear combinations of the drug indices with maximum variance, which suggests that a linear transformation must be performed among correlated variables, and the linearly transformed variables are subsequently arranged in order of decreasing variances 38 . A principal component is considered for inclusion when it has a loading eigenvalue of at least 1.0. For the retention of the drug property within a principal component, a minimum of 0.30 in absolute value is required as the correlation coefficient between the variable and its principal component. Sampling adequacy was assessed using Kaiser-Meyer-Olkin measure. Numerical calculations and data analysis were performed using STATA ® V12.0 software. After obtaining the principal components and identifying the significant indices within the components, additional clinically approved antiviral drugs including abacavir 39 , darunavir 40 , didanosine 41 , galidesivir 42 , stavudine 43 , and zalcitabine 44 for pharmacologic classification and some drugs that inhibited SARS-CoV-2 such as chloroquine 45 , lopinavir 45 , ivermectin 46 , and ciclesonide 45 were utilized to validate the results of the multivariate computational approach.

RESUlTS
Twenty different compounds with potential and clinically approved antiviral activity against SARS-CoV were classified using principal component analysis. Two principal components were obtained with eigenvalues 5.41 and 1.05. Within a component, the loading values of each drug index were computed ( Table 2). The important indices included in the first principal component were complexity (r = 0.4202), heavy atom count (r = 0.4123), and molecular weight (r = 0.4118) with a 77.35% contribution rate. The number of rotatable bond count (r = 0.7441) was the main index in the second principal component, with 14.97% contribution rate. The rotatable bond count, complexity, heavy atom count, and molecular weight indices primarily defined the pharmacologic classification of the compounds with potential and clinically approved antiviral activity (92.32% total contribution rate, 0.7620 Kaiser-Meyer-Olkin sampling adequacy). The main indices in the first principal component, namely complexity, heavy atom count, and Journal of Pure and Applied Microbiology molecular weight were positively correlated with the rest of the chemical and physical properties of the antiviral drugs (Table 3). Rotatable bond count, the leading index in the second principal component, was positively correlated with heavy atom count, complexity, and molecular weight ( Table 3).
The comprehensive scores for the different antiviral drugs were calculated using hierarchical weighted principal component analysis (range: 117.41-1087.30, Table 4). Interestingly, higher comprehensive scores (497.74-644.64) were identified among protease inhibitors (indinavir, nelfinavir, and saquinavir). In the validation of the computational approach for pharmacologic classification (Table  5), compounds such as abacavir, didanosine, galidesivir, stavudine, and zalcitabine obtained the lowest comprehensive scores (201.12-259.92). The comprehensive scores of these compounds were within the range of the comprehensive scores for nucleoside analogues and the RTIs (Table 4). Darunavir and lopinavir obtained comprehensive scores similar to those of other protease inhibitors examined in this study. The comprehensive score of ciclesonide was within the range of protease inhibitors, whereas ivermectin scored higher than any of the examined protease inhibitors but was inferior to glycyrrhizin and valinomycin. Moreover,

DISCUSSION
Among the chemical and physical properties of the potential and clinically approved antiviral drugs evaluated in this study, a positive correlation was identified in all drug properties, including complexity, heavy atom count, and molecular weight. Molecular weight has been considered an important compound property in small drug discovery 50 and is closely examined in drug optimization steps 51 . In addition to the Journal of Pure and Applied Microbiology molecular weight in the first principal component, the main index complexity was positively correlated with hydrogen bond acceptor and rotatable bond counts. Molecular complexity, which includes the cardinality of rings, stereocenters, and sp 3hybridized carbons, has been related to biological activity 52 . A compound with at least four aromatic rings has high toxicity risks and low compound developability 53 which justifies the preference for moderately complex structures as lead compounds 54 .
Compounds classified as NAIs, RTIs, protease inhibitors, nucleoside analogues, and some drugs with potential antiviral activity were examined. Among the nucleoside analogues, when acyclovir, foscavir, and ganciclovir were compared, there was an inverse relationship between the calculated comprehensive score and the biological activity (IC 50 ) of the antiviral drugs. For instance, when foscavir and acyclovir were examined against herpes simplex virus, acyclovir (IC 50 : 0.06 μmol/mL) was more potent than foscavir (IC 50 : 0.44 μmol/mL) 9 . The comprehensive score of acyclovir and foscavir was 207.22 and 124.11, respectively. Similarly, ganciclovir (IC 50 : 0.014 μmol/mL) was more potent than foscavir (IC 50 : 0.80 μmol/mL) against cytomegalovirus 9 . Ganciclovir had a higher comprehensive score (235.11) than foscavir (124.11). These observations led to the examination of a possible relationship between the comprehensive score based on physicochemical properties and the potency of the nucleoside analogues against viruses. The higher the comprehensive score of the nucleoside analogue, the more potent was the antiviral drug (lower IC 50 ). A nucleoside analogue inhibits viral polymerase and interferes with nucleic acid synthesis 18,24,25 . Acyclovir 18 , foscavir 24 , and ganciclovir 25 target DNA viruses such as varicellazoster virus and herpes simplex virus. However, acyclovir is more potent than foscavir (Foscarnet) as the former targets herpesvirus and varicellazoster virus polymerases 18 , whereas the latter selectively blocks the pyrophosphate binding site of herpes virus-specific DNA polymerases 24 . Among the nucleoside analogues examined in the present study, ganciclovir was the most potent (the lowest IC 50 and the highest comprehensive score). Ganciclovir inhibits replication of several viruses including varicella zoster virus, herpes simplex virus-1 and -2, Epstein-Barr virus, and cytomegalovirus 25 .
The fourth pharmacologic class of the antivirals examined was the NAIs. Representative drugs included were oseltamivir and zanamivir. Zanamivir had a higher comprehensive score than oseltamivir (334.50 vs. 267.02), suggesting that zanamivir is more potent than oseltamivir, and this comparison was supported by previous studies.
When the two NAIs were tested against influenza virus, zanamivir (Relenza) had a lower inhibitory concentration value than oseltamivir (Tamiflu) 9 . In addition, zanamivir (IC 50 : 2.7 nM) was significantly more potent than oseltamivir (IC 50 : 8.5 nM) when tested against influenza B virus isolates 57 .
Of the ten compounds utilized for validation, six were investigated for the pharmacologic classification and the remaining four for inhibitory potency. Validation of the computational approach for pharmacologic classification revealed low comprehensive scores among abacavir, didanosine, galidesivir, stavudine, and zalcitabine. These comprehensive scores were within the established ranges for nucleoside analogues and RTIs. Interestingly, abacavir, didanosine, and stavudine are nucleoside RTIs 39,41,43 . Abacavir has activity against HIV-1 (HIV-1IIIB EC 50 : 3.7-5.8 μM and HIV-1BaL EC 50 : 0.07-1.0 μM) 39 . Moreover, darunavir and lopinavir, both protease inhibitors 40,47 , had comprehensive scores similar to those of other protease inhibitors (indinavir, nelfinavir, and saquinavir). Similar to darunavir, lopinavir also inhibits the activity of an enzyme critical for the HIV viral lifecycle but has a high likelihood of drug interactions 47 .

CONClUSION
The chemical and physical properties of potential and clinically approved antiviral drugs explained their pharmacologic classification and biological activity. Hierarchically weighted principal component analysis elucidated the interaction between the physicochemical properties and SARS-CoV inhibition of these antiviral drugs. The physicochemical properties and inhibitory action against SARS-CoV-2 of lopinavir, chloroquine, ivermectin, and ciclesonide validated the adequacy of the current computational approach. The findings of the present study provide additional information, although further investigation is warranted to identify potential targets and establish exact mechanisms, in the emergent search and design of potential antiviral drug candidates and their subsequent synthesis as effective treatment against SARS-CoV-2 infection.