Virtual Screening of FDA-Approved Compounds: Exploring New Alternatives for HIV Treatment.

Introduction

The human immunodeficiency virus (HIV) is a retrovirus that affects approximately 39 million people worldwide. In 2023, approximately 630,000 global deaths were attributed to HIV infection. While still a significant concern, this number represents a 69% reduction in HIV-related deaths since the peak in 2004. This notable decline is primarily attributed to the widespread implementation of highly active antiretroviral therapy (HAART) for infection control. Estimates indicate that 77% of people living with HIV (PLHIV) were receiving HAART in 2023.

HAART’s primary objective is to diminish the morbidity and mortality associated with HIV infection. This objective is achieved by suppressing the viral load to undetectable levels, with viral suppression, PLHIV now experiences a life expectancy and quality of life comparable to those of individuals without the virus. Moreover, HAART and sustained viral suppression are crucial for reducing the risk of sexual transmission and preventing mother-to-child transmission. Currently, the concept of U = U (undetectable = untransmittable) exemplifies the profound success of HAART. Individuals with an undetectable viral load pose no risk of HIV transmission to their sexual partners. This understanding has significantly contributed to reducing the stigma associated with HIV and promoting effective treatment adherence among PLHIV.

Currently, over 30 FDA-approved drugs with diverse mechanisms of action are available for HIV treatment. These include nucleoside and nucleotide reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), protease (PIs) and integrase inhibitors (INIs), as well as fusion, attachment, postattachment, and capsid blockers, along with pharmacokinetic enhancers. According to the National Institutes of Health of the United States (NIH-USA), the recommended HAART regimen for treatment-naïve patients typically comprises a combination of two NRTIs alongside a third antiretroviral, wich can be an INI, a PI, or an NNRTI.

Despite these therapeutic advancements, two key challenges in HIV therapy demand ongoing attention: the emergence of drug-resistant strains and the side effects or adverse reactions associated with treatment. These factors frequently lead to poor patient adherence and, consequently, contribute to the development of resistance. After one year of HAART, researchers estimate that only 50–70% of patients maintain undetectable viral loads. Virological failure remains prevalent, with several associated risk factors.

Among the virus-related factors, the transmission of drug-resistant virus and the emergence of syncytium-inducing viral strains are particularly noteworthy. Therapy-related elements include the limited potency of HAART, poor penetration into sanctuary sites, and drug interactions that impair treatment effectiveness. In clinical care, the inadequate selection of drug combinations and the lack of HAART resistance testing are significant factors. Additionally, patient-related aspects such as poor adherence to therapy, genetic heterogeneity of the CCR5 and CXCR4 coreceptors, variations in the efficacy of intracellular phosphorylation of reverse transcriptase inhibitors, the presence of membrane proteins responsible for protease inhibitors efflux, gradual immune decline, and distinct patterns of HAART absorption and metabolism also influence treatment effectiveness.

Therefore, there is an ongoing need to identify new therapeutic candidates, ideally those that target novel mechanisms, offer improved pharmacokinetic profiles, and exhibit fewer side effects. Traditionally, high-throughput screening (HTS) has been the primary strategy for the discovery and evaluation of new molecules. While highly effective, this methodology is often time-consuming and costly for industry. Virtual screening, supported by computational methodologies such as the development of quantitative structure–activity relationship (QSAR) models and molecular docking, offers an approach to optimize this process, with several successful cases of chemoinformatics-driven drugs already reported. − Furthermore, computational strategies are also useful in predicting pharmacokinetic properties and drug-likeness, which are important parameters to assess in the early stages of new drug development.

Drug repositioning, a strategic approach in drug discovery, leverages virtual screening to identify new therapeutic uses for already approved molecules. This strategy is rooted in two key concepts: the idea that different diseases often share common biological targets, and the principle of pleiotropic drugs. Pleiotropic drugs are compounds that possess multiple therapeutic activities for multiple targets within a single molecule. , A significant advantage of drug repositioning is the streamlined regulatory process for market entry, as it already considers primary data on safety and toxicity. This effectively shortens the early development phases and leads to lower overall costs in bringing a new therapy to the market.

Given the possibility that different antiretrovirals may become ineffective in treating HIV over time, it is evident that exploring alternative approaches for developing active molecules against the virus is crucial. This includes both the prospecting of novel compounds and the repositioning of known molecules. The use of computational tools emerges as a promising solution to accelerate the discovery process, optimizing the identification of candidates with therapeutic potential while simultaneously reducing the costs and time associated with drug development. These strategies expand the range of treatment options available, which could positively impact on the quality of life of PLHIV.

Results

The best selected models, generated from PubChem bioassays, are summarized in Table . Models based on different algorithms, including decision trees, logistic regression, and gradient boostingamong others, were selected. The F1 scores ranged from 66.11% to 90.18%, while accuracy values ranged from 56.16% to 90.56%. All metrics were calculated using a balanced data set. The generated QSAR models were used to develop software called BAMBU-HIV, available at http://200.132.101.156:8085/ .

Among the best models, those with an accuracy exceeding 80% were selected for the subsequent stages. Of the 14 studies collected from PubChem, six exhibited an accuracy equal to or greater than 80%. Additionally, the results of the model generated for reverse transcriptase (PubChem AID: 565accuracy: 73.76%) were included due to the critical role of this target in the HIV viral replication process. Thus, in addition to reverse transcriptase, the best models developed for the Rev protein, glycoprotein gp160, GLS protein, glycoprotein gp41, integrase, and Nef protein were also utilized.

Using the models’ predictions, virtual screening was performed on a library of FDA-approved molecules. Based on the biological activity predictions performed by the QSAR models, the top 100 molecules from the library were selected for molecular docking, ranked according to their average QSAR prediction scores. This criterion was employed to prioritize molecules with a higher likelihood of exhibiting multitarget activity. The complete list of molecules is presented in Table SA of the Supporting Information .

To rank the molecular docking results, the top 20 molecules were selected based on their average docking energy. In the molecular docking assay, only molecules that demonstrated protein–ligand interactions with binding energies (Δ G ) ≤ −7.0 kcal/mol and relevant residues at the interface were considered. The complete list of molecular docking results is available in the Supporting Information , Table SB. The high-affinity targets most frequently identified among the results were protease (PDB: 5IVQ ) and reverse transcriptase (PDB: 3LP0 ). The molecule pioglitazone (ZINC000000968326) exhibited the highest number of interactions with various targets in the library, interacting with five distinct targets (PDB: 3F9K , 3LP0 , 5IVQ , 5UQE , 6B72 ). The binding energy ranged from −7.4 to −11.6 kcal/mol, the binding efficiency from −0.30 to −0.40, and the average PLIF score from 30.67% to 34.84%.

Table presents the toxicological profiles of the molecules analyzed using the ADMETLab 3.0 tool. This tool provides values ranging from 0 to 1, with low-toxicity compounds defined as those scoring between 0 and 0.3. Compounds with scores between 0.3 and 0.7 are considered to exhibit medium toxicity, while those above 0.7 are classified as highly toxic. In the neurotoxicity assessment, 10 molecules (10/20) were classified as highly toxic, nine (9/20) as low toxicity, and one (1/20) as medium toxicity. For ototoxicity, most molecules (13/20) exhibited high toxicity, and a similar trend was observed for nephrotoxicity (16/20), genotoxicity (15/20), hepatotoxicity (15/20), and respiratory toxicity (14/20).

Hematotoxicity was classified as low in 10 molecules (10/20), medium in seven (7/20), and high in three (3/20). Regarding carcinogenicity, 10 molecules were evaluated as having low carcinogenicity (10/20), seven as medium (7/20), and three (3/20) as high. Finally, in the evaluation of acute oral toxicity in an animal model, seven molecules (7/20) were considered low toxicity, another seven as medium (7/20), and six (6/20) as high toxicity. Icosapent ethyl (ZINC000003785276) exhibited the most favorable profile, presenting low toxicity in eight of the nine parameters evaluated and being classified as highly toxic only for respiratory toxicity. In contrast, riociguat (ZINC000003819392) was deemed highly toxic in eight of the nine parameters, with only acute oral toxicity in an animal model classified as medium.

Table presents the drug-likeness profile of the 20 molecules evaluated using the ADMETLab 3.0 tool. Lipinski’s rule stipulates that the molecular mass must be ≤500, the log P (octanol/water partition coefficient) must be ≤5, the number of hydrogen bond acceptors (Hacc) must be ≤10, and the number of hydrogen bond donors (Hdon) must be ≤5. According to the rule, for a drug to possess a favorable oral activity profile, it should not violate more than one of these parameters. Molecules that comply with Lipinski’s rule are classified as “excellent,” while those exhibiting two or more violations of the expected parameters are categorized as “poor.” Notably, all 20 evaluated molecules met the requirements of Lipinski’s rule (20/20).

According to the Pfizer rule, compounds with log P > 3 and a topological polar surface area (TPSA) < 75 are considered potentially toxic. Of the evaluated molecules, five (5/20) met both criteria and were classified as poor , while 15 molecules (15/20) met none or only one parameter, and were thus classified as excellent . The GSK rule stipulates that for compounds to exhibit a favorable ADMET profile, the molecular mass must be ≤400 and the log P ≤ 4. According to this rule, 11 (11/20) of the evaluated molecules were classified as excellent for meeting both criteria. Conversely, nine (9/20) did not satisfy at least one parameter and were classified as poor . Overall, ten molecules (10/20) were considered excellent across all rules utilized to assess drug-likeness in the tool. None of the top 20 molecules in the ranking list of the FDA-approved compound library exhibited PAINS-type interference.

Discussion and Conclusion

The use of ligand- and receptor-based virtual screening approaches enabled the identification of 20 FDA-approved molecules with potential for repurposing or repositioning. These molecules displayed affinity for various targets, and in some cases, multiple targets of interest for HIV.

Oncology Drugs

Pazopanib (ZINC000011617039) is an approved antineoplastic multikinase inhibitor used in renal cell carcinoma and soft tissue sarcoma, targeting VEGFR-1, -2, and -3, PDGFR-α/β, and FGFR-1 and -3, thereby inhibiting tumor angiogenesis. , In this study, pazopanib exhibited potential interaction with HIV protease (PLIF 35.28%). A retrospective analysis in PLHIV receiving oncological treatment reported that concomitant use of HAART with raltegravir and multikinase inhibitors did not compromise viral suppression, with undetectable viral loads and adequate plasma concentrations of the kinase inhibitors, suggesting a favorable safety profile for this combination.

To date, kinase inhibitors have primarily been employed in the treatment of cancer and inflammatory conditions. However, kinases are known to play significant roles in various physiological processes, making them promising pharmacological targets for further investigation. García-Cárceles reviewed the potential of kinase inhibitors as antiviral agents, emphasizing the connection between protein kinase inhibition and antiviral activity. Notably, they highlight dasatinib, an inhibitor that can suppress the activity of Src proteins, which are crucial for T cell activation, thereby creating an environment less conducive to HIV replication.

Despite these findings, pazopanib exhibited a high predicted toxicity profile in this study, consistent with its known risk of hepatotoxicity, which is highlighted in its clinical labeling. While current clinical practice recognizes that the benefits of HAART outweigh hepatotoxicity risks additional studies are required to further assess the safety and therapeutic feasibility of pazopanib in the context of HIV treatment.

Panobinostat (ZINC000051951647 and ZINC000022010649) is a nonselective histone deacetylase inhibitor approved for the treatment of multiple myeloma, acting through epigenetic modulation of gene expression and inhibition of protein metabolism. In the molecular docking assays, this drug demonstrated multiple potential interactions with reverse transcriptase, protease, and integrase (PLIF scores of 31.20% and 30.67%). Additionally, QSAR models suggested a possible interaction with gp41, although this was not detected in docking analyses.

When administered in conjunction with interferon-α2a, panobinostat appears to induce structural transformations in HIV-1 reservoir cells. The pharmacological induction of proviral gene expression may facilitate immunological clearance, aligning with the principles of “shock and kill” therapies designed to reduce HIV reservoirs. Armani-Tourret and colleagues demonstrated that the combined administration of panobinostat and interferon resulted in a decrease in nonclonal intact proviruses, a reduction in the structural and phylogenetic complexity of intact proviruses, and an increase in the number of proviruses with 5-LTR deletions. While such effects may occur naturally during long-term HAART, they appear to be accelerated by the described protocol, indicating that HIV-1 proviruses may be rendered and sensitized for immunological clearance. These findings are supported by clinical evidence demonstrating significant latency-reversing activity in PLHIV treated with panobinostat.

Additionally, the drug has been associated with reduced levels of plasma markers of inflammation, including C-reactive protein, interleukin-6 (IL-6), matrix metalloproteinase 9, and E-selectin in PLHIV. The suppression of these markers suggests a potential therapeutic benefit of panobinostat in PLHIV, particularly in individuals at high cardiovascular risk. Nevertheless, panobinostat exhibits a high toxicity profile, a common feature of oncological agents, underscoring the need for rigorous evaluation of the risk–benefit balance when considering its repurposing for HIV-related therapeutic strategies.

Medications for Pulmonary Conditions

Indacaterol (ZINC000035801098) and formoterol (ZINC000000897256 and ZINC000002599970) are long-acting β 2 -adrenergic agonists used for the management of asthma and chronic obstructive pulmonary disease by promoting bronchial muscle relaxation. Formoterol presents a comparatively more favorable toxicity profile, although both agents share common adverse effects such as nasopharyngitis and headache.

In this study, indacaterol (PLIF 34%) and formoterol (PLIF 32.13% and 32%) exhibited potential interactions with HIV reverse transcriptase and protease. QSAR models predicted high probabilities of interaction between indacaterol and the GLS protein (97.88%) and gp120 (72.91%), as well as between formoterol and GLS (97.96% and 96.53%), although these interactions were not supported by molecular docking analyses.

Evidence from the literature suggests potential antiviral activity of formoterol beyond its bronchodilator effects. In combination with budesonide, formoterol has been shown to reduce virus-induced inflammation by modulating pro-inflammatory cytokines and the type I interferon pathway, particularly in rhinovirus infections. Similar immunomodulatory effects have been reported in pediatric viral pneumonia, and drug repositioning studies have identified formoterol as a potential inhibitor of enterovirus A71 replication during early infection stages.

Riociguat (ZINC000003819392), a guanylate cyclase stimulator, is approved for the treatment of thromboembolic pulmonary hypertension and pulmonary arterial hypertension. The latter is a serious complication associated with HIV infection. DeJesus evaluated pharmacokinetic interactions of riociguat in patients undergoing HAART and reported an increase in the ROC curve of riociguat exposure in patients receiving the antiretroviral regimen abacavir/dolutegravir/lamivudine. Despite this pharmacokinetic alteration, the authors emphasized the overall safety of using riociguat in this patient population.

In this study, riociguat showed potential interaction with the GLS protein, which may be relevant in the context of HIV-associated dementia. However, the compound exhibited a high predicted toxicity profile, being classified as high risk in most evaluated parameters, and remains contraindicated during pregnancy due to teratogenic risk. These findings underscore the importance of a careful risk–benefit assessment before considering riociguat for broader therapeutic applications. To date, there are no reports describing interactions between riociguat and HIV molecular targets or its potential role in HIV-associated dementia, indicating that this remains an unexplored area that warrants further investigation.

Medications for Cardiovascular Conditions

Fluvastatin (ZINC000001886617) is a hydroxymethylglutaryl-coenzyme A reductase inhibitor widely used to reduce plasma cholesterol levels. Although generally well tolerated, fluvastatin is associated with hepatotoxicity, which is consistent with the toxicity profile predicted in this study. In molecular docking analyses, fluvastatin demonstrated potential interaction with HIV reverse transcriptase (PLIF 33.83%).

The addition of fluvastatin to the therapeutic regimen of patients coinfected with HIV and HCV has been associated with improved virological response. Moreover, statins exhibit immunomodulatory and anti-inflammatory properties, which are particularly relevant in the PLHIV population, characterized by chronic immune activation. Patients receiving HAART in combination with statins have shown reduced levels of inflammatory markers such as IP-10, IL-10, and IL-12p70.

Fluvastatin has also been reported to interfere with the replication of multiple enveloped and nonenveloped viruses, and other statins, such as lovastatin, have demonstrated the ability to reduce HIV viral load by disrupting host-cell signaling pathways involved in viral entry. , Additionally, statin use has been associated with decreased expression of host proteins involved in viral replication in SARS-CoV-2 infection. These antiviral effects are thought to be mediated, at least in part, by alterations in membrane lipid rafts, limiting cholesterol availability required for efficient viral fusion and gp41-mediated entry.

Ezetimibe (ZINC000003810860), indicated for primary and mixed hyperlipidemia, acts by inhibiting the sterol transporter NPC1L1 without interfering with the absorption of fat-soluble nutrients. Based on the results obtained in this study, the toxicity profile of ezetimibe can be classified as medium to high. Ezetimibe demonstrated potential interactions with reverse transcriptase and protease (PLIF 31.11%). Although QSAR models predicted potential interactions with integrase (78.14%) and the GLS protein (97.82%), these were not confirmed through molecular docking. Preliminary studies have suggested an antiviral effect of ezetimibe against various viruses, including hepatitis B virus hepatitis C virus, as well as dengue, zika, and yellow fever virus.

Icosapent ethyl (ZINC000003785276), an ethyl ester of eicosapentaenoic acid, is approved for the treatment of severe hypertriglyceridemia and is characterized by a favorable safety profile. Among the compounds evaluated, it showed the lowest predicted toxicity and exhibited potential interactions with HIV integrase and reverse transcriptase in docking analyses (PLIF 32.43%), and a strong probability of interaction with integrase was predicted in a model derived from bioassay 1053136 (99.99%). To date, there are no documented interactions in the literature between ezetimibe and icosapent ethyl with the targets identified in this study.

Despite its safety profile, the immunomodulatory effects of polyunsaturated fatty acids such as icosapent ethyl warrant caution. While supplementation may be beneficial in infections marked by excessive inflammation, prolonged use can alter gut microbiota and dampen immune responses, potentially compromising host defense in certain infectious contexts.

Labetalol (ZINC000000403011), is an α- and β-adrenergic antagonist approved for the treatment of hypertension and related cardiovascular conditions, acting primarily through vasodilation. In this study, labetalol exhibited potential multitarget interactions with HIV reverse transcriptase, protease, GLS protein, and Nef protein (PLIF 31.56%). Although QSAR models predicted a strong interaction with the GLS protein, a high-probability interaction with integrase was not supported by molecular docking results. Despite the probable multitarget interactions identified in this study, there are no prior reports in the literature documenting labetalol’s interactions with the evaluated receptors. Its toxicity profile, as determined in this study, is considered high, particularly regarding nephrotoxicity and hepatotoxicity.

Medications from Diverse Classes

Conivaptan (ZINC00001250318) is a vasopressin receptor isotype inhibitor indicated for the treatment of hyponatremia. In molecular docking assays, it demonstrated a potential interaction with protease (PLIF 34.84%). Although QSAR models suggested a possible interaction with reverse transcriptase, this finding was not supported by docking analyses. Notably, virtual screening studies have identified conivaptan as a potential antiviral candidate against mpox virus and SARS-CoV-2, supporting its broader antiviral relevance. ,

Latanoprost (ZINC000012468792) is a prostaglandin F2α analogue prodrug widely used for the treatment of elevated intraocular pressure, with minimal systemic adverse effects. In this study, latanoprost demonstrated potential interaction with HIV reverse transcriptase (PLIF 32.47%). While QSAR models predicted a high probability of interaction with integrase, this result was not confirmed by molecular docking.

Calcitriol (ZINC000100015048), the active metabolite of vitamin D, is widely used in the management of secondary hyperparathyroidism and psoriasis. , In molecular docking assays, calcitriol exhibited a potential interaction with protease (PLIF 31.91%). The QSAR model suggested a possible interaction with integrase (99.85%), which was not confirmed through docking. Evidence suggests that vitamin D and its metabolites may exert beneficial effects in PLHIV through immunomodulatory mechanisms. Vitamin D has been associated with reduced HIV-1 infection in T cells, decreased chronic inflammation, and improved immunological outcomes, including increased CD4 T cell counts and reduced viral load, supporting its potential role as an adjunct to HAART. , ,

Droperidol (ZINC000019796080) is a dopamine antagonist derived from butyrophenone, used to prevent and treat nausea and vomiting in postsurgical patients. In molecular docking assays, it exhibited multiple potential interactions with integrase, reverse transcriptase, protease, and the GLS protein (PLIF 31.89%).

Ziprasidone (ZINC000000538550) is an atypical antipsychotic used in the treatment of schizophrenia and bipolar disorder and is also prescribed for HIV/AIDS-related psychosis. As a CYP3A4 substrate, its plasma concentrations may increase when coadministered with protease inhibitors such as ritonavir, and its potential to prolong the QT interval requires caution, particularly in combinations with other QT-prolonging agents. In molecular docking assays, ziprasidone demonstrated potential interactions with HIV protease and the GLS protein (PLIF 31.76%).

Cilastatin (ZINC000004095696) demonstrated potential interactions with protease and reverse transcriptase (PLIF 31.34%). QSAR models predicted possible interactions with integrase and the GLS protein, which were not confirmed in molecular docking assays. Cilastatin functions as an inhibitor of renal dehydropeptidase, an enzyme responsible for metabolizing β-lactam antibiotics. It is clinically used in combination with imipenem to prevent the metabolism of the antibiotic.

Dinoprostone (ZINC000003830713), a naturally occurring prostaglandin E 2 , is clinically used for labor induction and related obstetric indications. In this study, dinoprostone demonstrated potential interaction with HIV reverse transcriptase in molecular docking assays (PLIF 31.06%). Although QSAR models predicted interactions with integrase, GLS protein, gp160, and gp120, these were not confirmed by docking. Notably, prostaglandin E 2 has been implicated in HIV replication dynamics, and postinfection exposure has been shown to reduce viral replication by impairing the intracellular transport and assembly of viral components, thereby decreasing virion release.

Pioglitazone (ZINC000000968326) exhibited potential multitarget interactions with integrase, reverse transcriptase, protease, GLS protein, and Nef protein in molecular docking assays (PLIF 30.87%). Notably, none of these interactions were predicted with high probability (>70%) by QSAR models. Originally developed as an antihyperglycemic agent for type 2 diabetes mellitus, pioglitazone has fallen into disuse over the years due to multiple adverse effects, including congestive heart failure and bladder cancer. Its use in PLHIV is particularly associated with metabolic-associated fatty liver disease, a complication in this population that can progress to cirrhosis.

In a study with 98 patients receiving HAART and pioglitazone for 48 weeks, no serious adverse effects were reported, and the drug significantly improved parameters associated with fatty liver disease. Additionally, pioglitazone has shown broad in vitro antiviral activity against SARS-CoV-2 variants when combined with pamapimod.

Finally, cefprozil (ZINC000004474443), a cephalosporin-class antibiotic, showed a potential interaction with protease (PLIF 30.86%). QSAR models predicted additional high-probability interactions with integrase and GLS protein; however, they were not confirmed by molecular docking. Among the drugs classified in the “diverse classes” section, few reports exist in the literature regarding interactions with the HIV targets discussed here, except for vitamin D derivatives and dinoprostone, which require further robust studies to elucidate their mechanisms of action.

Regarding the toxicity profiles of the various compounds, latanoprost, cilastatin, dinoprostone, pioglitazone, and cefprozil demonstrated the most favorable outcomes across the nine evaluated parameters. Conivaptan and droperidol exhibited medium to high toxicity profiles, particularly concerning genotoxicity, with results approaching the maximum limit. Calcitriol and ziprasidone were classified as having high toxicity across most parameters.

Although all compounds evaluated are FDA-approved, adverse effects remain an important consideration, especially in PLHIV, who require lifelong combination therapy. Therefore, toxicity assessment is critical when considering drug repositioning, particularly with respect to potential interactions with existing HAART regimens, as monotherapy is not recommended.

None of the compounds violated Lipinski’s rule of five. However, violations were more frequent under the more restrictive GSK rule, particularly among icosapent ethyl, calcitriol, ziprasidone, and ezetimibe, which also violated the Pfizer rule. Since the GSK rule imposes stricter limits on molecular weight and lipophilicity, such violations may affect bioavailability and tissue distribution. Additionally, most compounds violating the Pfizer rule were also classified as highly toxic, reinforcing the importance of integrating drug-likeness and toxicity metrics in future drug repositioning strategies.

QSAR models and molecular docking assays evaluate potential interactions between molecules and target proteins using distinct methodologies. This fundamental difference explains why many interactions predicted with high probability by QSAR models were not corroborated by molecular docking assays. , QSAR relies on molecular descriptors that capture the topological and physiochemical properties of the analyzed compounds, with the correlation between the descriptors and the expected experimental activity modeled through statistical or machine learning models. However, it is well recognized that QSAR models are highly dependent on the nature of their training set. They tend to exhibit strong predictive performance when applied to compounds that are structurally related to those in the training set, but their predictive power decreases when they are used to evaluate structurally diverse molecules, as is the case in the present study.

In contrast, molecular docking does not require prior ligand activity data but depends on accurate structural information on the target protein. Although docking scoring functions often show limited quantitative correlation with experimental binding affinities, they are generally effective in discriminating between active and inactive compounds by estimating relative binding energies rather than absolute interaction energies. Taken together, these findings underscore the complementary nature of ligand-based and structure-based approaches and support their combined application in virtual screening pipelines.

These discrepancies between the results obtained from QSAR models and molecular docking highlight the need for caution when interpreting computational findings. It is important to recognize that virtual screening outcomes are directly affected by errors inherent in the input data, which propagate through the constructed models and consequently influence their predictions. Nonetheless, in high-throughput screening assays, the expected hit rate is typically below 1%, which supports the use of computational approaches as an alternative to enhance predictive success rates. Additionally, approximately 39% of the molecules evaluated in this study were confirmed by the PASS Online software as having potential activity against HIV-1, HIV-2, or exhibiting general antiviral properties.

By employing virtual screening, some molecules identified as promising hits may be commercially unavailable, either because they have been discontinued by the pharmaceutical industry, withdrawn from the market due to regulatory issues, or retained only for highly specific applications. This does not appear to be the case for the hits identified in the present study, as all molecules are commercially available, and because we chose to use a library of molecules approved by the USA regulatory agency (FDA).

Drug repurposing substantially reduces the risks and costs associated with drug development, as key stages such as preclinical evaluation, toxicological testing, chemical optimization, and formulation have already been completed. As a result, the development timeline can be reduced from approximately 15 years to about 6.5 years, with estimated costs decreasing from 2 to 3 billion to around 300 million United States dollars. In line with this, repurposed drugs exhibit a lower overall probability of failure compared to de novo drug discovery approaches.

Despite these advantages, drug repositioning faces important challenges, particularly when compounds remain under patent protection, which limits access to safety and efficacy data. Similar constraints apply to discontinued or abandoned drug candidates that progressed through early clinical development but failed to achieve regulatory approval due to insufficient efficacy, safety concerns, or financial limitations. Given that nearly 90% of drug candidates entering clinical trials do not reach approval, restricted data access often makes repositioning efforts more feasible for patent-holding pharmaceutical companies than for academic or small biotechnology groups.

Conclusions

Regarding the drugs evaluated, it is important to emphasize that these results represent preliminary findings. Nonetheless, this study, combined with existing literature, highlights pazopanib, panobinostat, fluvastatin, and calcitriol as promising candidates in this virtual screening approach. Further research is essential to explore and validate these molecular interactions and assess their clinical relevance. The QSAR models developed alongside molecular docking assays have provided valuable insights into potential novel compounds for HIV treatment. Based on these models, a software tool has been created to facilitate the academic community’s screening of additional molecules.

Utilizing a computational approach to prioritize molecules, particularly those with multitarget potential, significantly accelerates the drug discovery process for complex diseases like HIV. However, this also highlights the critical need for robust experimental studies to validate the predictive findings and confirm the efficacy of the identified compounds.

Materials and Methods

Data Search and Selection

The study began with the search and selection of data from high-throughput screening (HTS) studies available in the PubChem database ( https://pubchem.ncbi.nlm.nih.gov/ ) to train and validate QSAR models. PubChem is an open-access chemical database that provides extensive information on chemical structures, biological activities, safety, and toxicity. The “bioassays” entry was used for data search, using “HIV” as the keyword and focusing on studies aimed at developing new HIV drugs. Only bioassays containing at least 500 active molecules were included to ensure a sufficient data set for both model training and validation.

The BAMBU tool was used to train and validate the QSAR models. BAMBU is a command-line software that autogenerates QSAR models using HTS data as input. Bioassay data were downloaded from PubChem via the bambu-download command, which generates a CSV file containing molecular information. Each molecule is represented by its InChl strings in the “InChl” column and labeled as either “active” or “inactive” in the “activity” column.

Table summarizes the selected studies, listing the AID (PubChem assay identification number), study targets, total number of compounds tested, and the distribution between active and inactive compounds. While most studies focused on discovering new HIV drugs, some targeted glycoprotein gp160 and sialate acetyl esterase, aiming to identify compounds relevant for vaccine development. These targets were also included due to their biological significance and the potential value of the compound libraries. Based on the established criteria, 14 studies were selected for QSAR model training and validation.

Training and Validation of QSAR Models

For model development, the data were initially downloaded and preprocessed using the bambu-preprocess command. At this stage, the raw data set was partitioned into two subsets: 75% for training and 25% for validation. Molecules, originally in InChI format, were subsequently converted into different representations. BAMBU uses the MORDRED package to compute over 1200 structural and physicochemical molecular descriptors (both 2D and 3D). Additionally, models were generated employing Morgan fingerprints (1024 or 2048 bits) and vector representations via Mol2Vec. To mitigate the significant class imbalance in the data set, a random subsampling technique was applied. This process utilized the following packages: Pandas ( https://pandas.pydata.org/ ), NumPy ( https://numpy.org/ ), Scikit-learn ( https://scikit-learn.org ) and Imbalanced-learn ( https://imbalanced-learn.org/ ).

Following preprocessing, the models were trained using the bambu-train command. This training process utilized the subset of training data to identify the best classification model through the application of optimized hyperparameters. BAMBU incorporates fast and lightweight automated machine learning (FLAML), a machine learning framework developed by Microsoft, which automatically selects the most suitable model and hyperparameters for each study. FLAML’s methodology includes algorithms such as random forest, gradient boosting, logistic regression, and extra-trees.

Once the optimal hyperparameter configuration was selected, the predictive models were refined using the training data. The bambu-validate command was employed to perform validation, calculating accuracy, recall, precision, F1 score, and the area under the ROC curve (AUC), with AUC serving as the primary evaluation metric. Based on the classification metrics derived from both the training and test sets, p-values were calculated for each metric. An additional step to evaluate the selected model involved k-fold cross-validation ( k = 10) on the training data set. In this process, the data set was partitioned into k partitions, and in each iteration, k – 1 partitions were used for training, while the remaining subset was used for validation. Figure illustrates the BAMBU workflow in detail.

Virtual Screening

Virtual screening encompasses the ligand-based design or repositioning step. For this process, the best QSAR models developed for each target were employed. Molecules from the Zinc15 database ( https://zinc15.docking.org/ ), specifically from the “ FDA Approved ” library (comprising 1.615 drugs), underwent an initial filtering. Subsequently, an average score was computed for these molecules based on the results generated by the different models. The top 1.000 promising molecules were then selected for the molecular docking phase.

Molecular Docking

In the molecular docking phase, simulations of docking between proteins and small ligands were performed using AutoDock Vina, which employs its standard empirical scoring function, based on ligand–receptor interaction energy terms. No custom modifications were made to the weights or individual components of the scoring function. Ligand and receptor structures were prepared using AutoDock Tools. HIV target structures were obtained from the Protein Data Bank (PDB), and relevant residues for each target were identified through a comprehensive literature review, including prior studies detailing their mechanisms of action or molecular docking investigations. When critical residues were not found in the literature, the active site predictor PrankWeb was employed. Additionally, the protease structure was incorporated in the docking process due to its critical role in HIV, despite the absence of a generated QSAR model, as no HTS data were available for this structure in PubChem. The identified residues are summarized in Table , available in the Supporting Information .

For each target structure, a grid box of 10 Å in all three axes was generated around the selected residues. The docking results were subsequently processed using the PLIP tool, which calculated binding interaction scores for each protein–ligand complex. These scores were derived from the percentage of relevant residues, identified in the literature review, that exhibited predicted interactions with a specific ligand.

Protein–ligand interaction fingerprints (PLIFs) were then utilized to compute an average score across all targets, enabling the ranking of molecules. High-affinity targets were defined as those exhibiting binding energies ≥−7.0 kcal/mol and possessing at least 0.3 heavy atoms in the ligand. , Only molecules with at least one high-affinity target were selected for the subsequent phase.

Analysis of Pharmacokinetic Characteristics, Toxicity, and Promiscuous Compounds

The pharmacokinetic and toxicological properties, alongside structural alerts for molecular promiscuity, were evaluated for each molecule using the ADMETLab 3.0 web server Following these comprehensive analyses, a final list of the 20 most promising molecules from the analyzed library was compiled.

Supplementary Material

The authors acknowledge the Technological Development Center (CDTec) at the Federal University of Pelotas (UFPel) for technical support. This study was supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, Brasil).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c06562 . Table SAlist of the 100 molecules from the FDA-approved compound library selected for molecular docking; and Table SBtop 20 molecules from the FDA-approved drug library with high-affinity targets identified by molecular docking for different HIV-related targets. Binding energy and binding efficiency are reported in kcal/mol. The protein–ligand interaction fingerprint was compared with residues previously described in the literature or predicted using specific tools, which were considered relevant for the activity of molecules targeting the proteins ( PDF )

D.P.M. and F.S.K. contributed equally. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

This work was carried out with financial support from the Brazilian Federal Agency for Support and Evaluation of Graduate Education (CAPES), Brazil. The Article Processing Charge for the publication of this research was funded by the Coordenacao de Aperfeicoamento de Pessoal de Nivel Superior (CAPES), Brazil (ROR identifier: 00x0ma614).

The authors declare no competing financial interest.

UNAIDS/WHO . HIV Data and Statistics , 2023 . Available from: https://www.who.int/teams/global-hiv-hepatitis-and-stis-programmes/hiv/strategic-information/hiv-data-and-statistics .

The Antiretroviral Therapy Cohort Collaboration Survival of HIV-positive patients starting antiretroviral therapy between 1996 and 2013: a collaborative analysis of cohort studies Lancet HIV 2017 4 8 e349 e356 10.1016/S2352-3018(17)30066-8 28501495 PMC5555438

Quinn T. C. Wawer M. J. Sewankambo N. Serwadda D. Li C. Wabwire-Mangen F. Meehan M. O. Lutalo T. Gray R. H. Viral load and heterosexual transmission of human immunodeficiency virus type 1 N. Engl. J. Med. 2000 342 13 921 929 10.1056/nejm200003303421303 10738050

Eisinger R. W. Dieffenbach C. W. Fauci A. S. HIV Viral Load and Transmissibility of HIV Infection: Undetectable Equals Untransmittable JAMA 2019 321 5 451 452 10.1001/jama.2018.21167 30629090

Department of Health and Human Services . Guidelines for the Use of Antiretroviral Agents in Adults and Adolescents with HIV , 2023 .

Zhan P. Pannecouque C. De Clercq É. Liu X. Anti-HIV Drug Discovery and Development: Current Innovations and Future Trends J. Med. Chem. 2016 59 7 2849 2878 10.1021/acs.jmedchem.5b00497 26509831

Blaise P. Clevenbergh P. Vaira D. Moutschen M. Dellamonica P. HIV Resistance to Antiretroviral Drugs: Mechanisms, Genotypic and Phenotypic Resistance Testing in Clinical Practice Acta Clin. Belg. 2002 57 4 191 201 10.1179/acb.2002.041 12462795

Macarron R. Banks M. N. Bojanic D. Burns D. J. Cirovic D. A. Garyantes T. Green D. V. S. Hertzberg R. P. Janzen W. P. Paslay J. W. Impact of high-throughput screening in biomedical research Nat. Rev. Drug Discov. 2011 10 3 188 195 10.1038/nrd3368 21358738

Muegge I. Bergner A. Kriegl J. M. Computer-aided drug design at Boehringer Ingelheim J. Comput. Aided Mol. Des. 2017 31 275 285 10.1007/s10822-016-9975-3 27650777

Schneider G. Virtual screening: an endless staircase? Nat. Rev. Drug Discov. 2010 9 4 273 276 10.1038/nrd3139 20357802

Zhou S. F. Zhong W. Z. Drug Design and Discovery: Principles and Applications Molecules 2017 22 2 279 10.3390/molecules22020279 28208821 PMC6155886

Norinder U. Bergström C. A. S. Prediction of ADMET Properties ChemMedChem 2006 1 9 920 937 10.1002/cmdc.200600155 16952133

Guiselin T. Lecoutey C. Rochais C. Dallemagne P. Conceptual Framework of the Design of Pleiotropic Drugs against Alzheimer’s Disease Pharmaceutics 2023 15 10 2382 10.3390/pharmaceutics15102382 37896142 PMC10610275

Jourdan J. Bureau R. Rochais C. Dallemagne P. Drug repositioning: a brief overview J. Pharm. Pharmacol. 2020 72 9 1145 1151 10.1111/jphp.13273 32301512 PMC7262062

Benet L. Z. Hosey C. Ursu O. Oprea T. BDDCS, the Rule of 5 and drugability Adv. Drug Deliv. Rev. 2016 101 89 98 10.1016/j.addr.2016.05.007 27182629 PMC4910824

Hughes J. D. Blagg J. Price D. Bailey S. Decrescenzo G. Devraj R. Physiochemical drug properties associated with in vivo toxicological outcomes Bioorg. Med. Chem. Lett. 2008 18 4872 4875 10.1016/j.bmcl.2008.07.071 18691886

Gleeson M. P. Generation of a set of simple, interpretable ADMET rules of thumb J. Med. Chem. 2008 51 817 834 10.1021/jm701122q 18232648

Miyamoto S. Kakutani S. Sato Y. Hanashi A. Kinoshita Y. Ishikawa A. Drug review: Pazopanib Jpn. J. Clin. Oncol. 2018 48 6 503 513 10.1093/jjco/hyy053 29684209

Sonpavde G. Hutson T. E. Sternberg C. N. Pazopanib, a potent orally administered small-molecule multitargeted tyrosine kinase inhibitor for renal cell carcinoma Expert Opin. Invest. Drugs 2008 17 2 253 261 10.1517/13543784.17.2.253 18230058

Loulergue P. Merad M. Coriat R. Ducreux M. Planchard D. Boige V. Le Cesne A. Gregory T. M. Poinsignon V. Paci A. Safety of raltegravir-based antiretroviral therapy in HIV-infected patients receiving multi-kinase inhibitors Invest. New Drugs 2017 35 2 247 249 10.1007/s10637-016-0405-0 27838867

García-Cárceles J. Caballero E. Gil C. Martínez A. Kinase Inhibitors as Underexplored Antiviral Agents J. Med. Chem. 2022 65 2 935 954 10.1021/acs.jmedchem.1c00302 33970631 PMC8802305

Boudou-Rouquette P. Tlemsani C. Blanchet B. Huillard O. Jouinot A. Arrondeau J. Thomas-Schoemann A. Vidal M. Alexandre J. Goldwasser F. Clinical pharmacology, drug-drug interactions and safety of pazopanib: a review Expert Opin. Drug Metab. Toxicol. 2016 12 12 1433 1444 10.1080/17425255.2016.1225038 27556889

Antonello V. S. Kliemann D. A. Santos B. R. Tovo C. V. HAART and liver: is it safe? J. Infect. Dev. Countries 2014 8 11 1444 1450 10.3855/jidc.5012 25390056

Qian D. Z. Kato Y. Shabbeer S. Wei Y. Verheul H. M. W. Salumbides B. Sanni T. Atadja P. Pili R. Targeting tumor angiogenesis with histone deacetylase inhibitors: the hydroxamic acid derivative LBH589 Clin. Cancer Res. 2006 12 2 634 642 10.1158/1078-0432.ccr-05-1132 16428510

Armani-Tourret M. Gao C. Hartana C. A. Sun W. Carrere L. Vela L. Hochroth A. Bellefroid M. Sbrolla A. Shea K. Selection of epigenetically privileged HIV-1 proviruses during treatment with panobinostat and interferon-α2a Cell 2024 187 5 1238 1254.e14 10.1016/j.cell.2024.01.037 38367616 PMC10903630

Olesen R. Vigano S. Rasmussen T. A. Søgaard O. S. Ouyang Z. Buzon M. Bashirova A. Carrington M. Palmer S. Brinkmann C. R. Innate Immune Activity Correlates with CD4 T Cell-Associated HIV-1 DNA Decline during Latency-Reversing Treatment with Panobinostat J. Virol. 2015 89 20 10176 10189 10.1128/jvi.01484-15 26223643 PMC4580197

Høgh Kølbæk Kjær A. S. Brinkmann C. R. Dinarello C. A. Olesen R. Østergaard L. Søgaard O. S. Tolstrup M. Rasmussen T. A. The histone deacetylase inhibitor panobinostat lowers biomarkers of cardiovascular risk and inflammation in HIV patients AIDS 2015 29 10 1195 1200 10.1097/qad.0000000000000678 25870990

Naline E. Trifilieff A. Fairhurst R. A. Effect of indacaterol, a novel long-acting 2-agonist, on isolated human bronchi Eur. Respir. J. 2007 29 3 575 581 10.1183/09031936.00032806 17135231

Davies S. J. Burhan A. M. Kim D. Gerretsen P. Graff-Guerrero A. Woo V. L. Sequential drug treatment algorithm for agitation and aggression in Alzheimer’s and mixed dementia J. Psychopharmacol. 2018 32 5 509 523 10.1177/0269881117744996 29338602 PMC5944080

Lin Z.-L. Xu X. Yang J. J. Lu L. Q. Huang H. Hua X. Z. Lu L. D. Efficacy of budesonide/formoterol inhalation powder in treating viral pneumonia in children World J. Clin. Cases 2024 12 21 4469 4475 10.12998/wjcc.v12.i21.4469 39070822 PMC11235498

Jintana K. Prasertsopon J. Puthavathana P. Lerdsamran H. Antiviral effect in association with anti-apoptosis and anti-autophagy of repurposing formoterol fumarate dihydrate on enterovirus A71-infected neuronal cells Virus Res. 2022 311 198692 10.1016/j.virusres.2022.198692 35093474

DeJesus E. Saleh S. Cheng S. Van der Mey D. Becker C. Frey R. Pharmacokinetic interaction of riociguat and antiretroviral combination regimens in HIV-1-infected adults Pulm. Circ. 2019 9 2045894019848644 10.1177/2045894019848644 30997864 PMC6540510

Riociguat: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB08931 (cited Jun 27, 2025 ).

Milazzo L. Caramma I. Mazzali C. Cesari M. Olivetti M. Galli M. Antinori S. Fluvastatin as an adjuvant to pegylated interferon and ribavirin in HIV/hepatitis C virus genotype 1 co-infected patients: an open-label randomized controlled study J. Antimicrob. Chemother. 2010 65 4 735 740 10.1093/jac/dkq002 20118492

Hussain S. K. Golozar A. Widney D. P. Rappocciolo G. Penugonda S. Bream J. H. Martínez-Maza O. Jacobson L. P. Effect of Statin Use on Inflammation and Immune Activation Biomarkers in HIV-Infected Persons on Effective Antiretroviral Therapy AIDS Res. Hum. Retroviruses 2021 37 5 357 367 10.1089/aid.2020.0127 33238713 PMC8112711

Mohd Jaafar F. Monsion B. Belhouchet M. Mertens P. P. C. Attoui H. Inhibition of Orbivirus Replication by Fluvastatin and Identification of the Key Elements of the Mevalonate Pathway Involved Viruses 2021 13 8 1437 10.3390/v13081437 34452303 PMC8402872

del Real G. Jiménez-Baranda S. Mira E. Lacalle R. A. Lucas P. Gómez-Moutón C. Alegret M. Peña J. M. Rodríguez-Zapata M. Alvarez-Mon M. Statins inhibit HIV-1 infection by down-regulating Rho activity J. Exp. Med. 2004 200 4 541 547 10.1084/jem.20040061 15314078 PMC2211926

Zapatero-Belinchón F. J. Moeller R. Lasswitz L. van Ham M. Becker M. Brogden G. Fluvastatin mitigates SARS-CoV-2 infection in human lung cells iScience 2021 24 12 103469 10.1016/j.isci.2021.103469 34812415 PMC8599137

Gorabi A. M. Kiaie N. Bianconi V. Jamialahmadi T. Al-Rasadi K. Johnston T. P. Pirro M. Sahebkar A. Antiviral effects of statins Prog. Lipid Res. 2020 79 101054 10.1016/j.plipres.2020.101054 32777243

Ezetimibe: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00973 (cited Jun 27, 2025 ).

Lucifora J. Esser K. Protzer U. Ezetimibe blocks hepatitis B virus infection after virus uptake into hepatocytes Antiviral Res. 2013 97 2 195 197 10.1016/j.antiviral.2012.12.008 23266293

Feld J. J. Cypel M. Kumar D. Dahari H. Pinto Ribeiro R. V. Marks N. Kamkar N. Bahinskaya I. Onofrio F. Q. Zahoor M. A. Short-course, direct-acting antivirals and ezetimibe to prevent HCV infection in recipients of organs from HCV-infected donors: a phase 3, single-centre, open-label study Lancet Gastroenterol. Hepatol. 2020 5 7 649 657 10.1016/s2468-1253(20)30081-9 32389183 PMC7391837

Osuna-Ramos J. F. Farfan-Morales C. N. Cordero-Rivera C. D. De Jesús-González L. A. Reyes-Ruiz J. M. Hurtado-Monzón A. M. Palacios-Rápalo S. N. Jiménez-Camacho R. Meraz-Ríos M. A. Del Ángel R. M. Cholesterol-Lowering Drugs as Potential Antivirals: A Repurposing Approach against Flavivirus Infections Viruses 2023 15 7 1465 10.3390/v15071465 37515153 PMC10383882

Ballantyne C. M. Braeckman R. A. Soni P. N. Icosapent ethyl for the treatment of hypertriglyceridemia Expert Opin. Pharmacother. 2013 14 10 1409 1416 10.1517/14656566.2013.798645 23701295

Husson M.-O. Ley D. Portal C. Gottrand M. Hueso T. Desseyn J. L. Gottrand F. Modulation of host defence against bacterial and viral infections by omega-3 polyunsaturated fatty acids J. Infect. 2016 73 6 523 535 10.1016/j.jinf.2016.10.001 27746159

Labetalol: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00598 (cited Jun 27, 2025 ).

Mao Z. L. Stalker D. Keirns J. Pharmacokinetics of conivaptan hydrochloride, a vasopressin V­(1A)/V(2)-receptor antagonist, in patients with euvolemic or hypervolemic hyponatremia and with or without congestive heart failure from a prospective, 4-day open-label study Clin. Ther. 2009 31 7 1542 1550 10.1016/j.clinthera.2009.07.011 19695403

Dutt M. Kumar A. Rout M. Dehury B. Martinez G. Ndishimye P. Kelvin A. A. Kelvin D. J. Drug repurposing for Mpox: Discovery of small molecules as potential inhibitors against DNA-dependent RNA polymerase using molecular modeling approach J. Cell. Biochem. 2023 124 5 701 715 10.1002/jcb.30397 36946432 PMC10473176

Gul S. Ozcan O. Asar S. Okyar A. Barıs I. Kavakli I. H. In silico identification of widely used and well-tolerated drugs as potential SARS-CoV-2 3C-like protease and viral RNA-dependent RNA polymerase inhibitors for direct use in clinical trials J. Biomol. Struct. Dyn. 2021 39 17 6772 6791 10.1080/07391102.2020.1802346 32752938 PMC7484590

Latanoprost: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00654 (cited Jun 27, 2025 ).

Aguilar-Jimenez W. Villegas-Ospina S. Gonzalez S. Zapata W. Saulle I. Garziano M. Biasin M. Clerici M. Rugeles M. T. Precursor Forms of Vitamin D Reduce HIV-1 Infection In Vitro J. Acquir. Immune Defic. Syndr. 2016 73 5 497 506 10.1097/qai.0000000000001150 27509245

Calcitriol: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00136 (cited Jun 27, 2025 ).

Alvarez N. Aguilar-Jimenez W. Rugeles M. T. The Potential Protective Role of Vitamin D Supplementation on HIV-1 Infection Front. Immunol. 2019 10 2291 10.3389/fimmu.2019.02291 31611877 PMC6773828

Qurban R. Saeed S. Kanwal W. Junaid K. Rehman A. Potential immune modulatory effect of vitamin D in HIV infection: A review Clin. Nutr. ESPEN 2022 47 1 8 10.1016/j.clnesp.2021.12.005 35063189

Droperidol: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00450 (cited Jun 27, 2025 ).

Stahl S. M. Shayegan D. K. The psychopharmacology of ziprasidone: receptor-binding properties and real-world psychiatric practice J. Clin. Psychiatry 2003 64 Suppl. 19 6 12 14728084

Hill L. Lee K. C. Pharmacotherapy considerations in patients with HIV and psychiatric disorders: focus on antidepressants and antipsychotics Ann. Pharmacother. 2013 47 1 75 89 10.1345/aph.1r343 23341158

Keynan S. Hooper N. M. Felici A. Amicosante G. Turner A. J. The renal membrane dipeptidase (dehydropeptidase I) inhibitor, cilastatin, inhibits the bacterial metallo-beta-lactamase enzyme CphA Antimicrob. Agents Chemother. 1995 39 7 1629 1631 10.1128/aac.39.7.1629 7492120 PMC162797

Dinoprostone: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB00917 (cited Jun 27, 2025 ).

Clemente M. I. Álvarez S. Serramía M. J. Martínez-Bonet M. Muñoz-Fernández M. Á. Prostaglandin E2 reduces the release and infectivity of new cell-free virions and cell-to-cell HIV-1 transfer PLoS One 2014 9 2 e85230 10.1371/journal.pone.0085230 24586238 PMC3934822

Matthews L. Kleiner D. E. Chairez C. McManus M. Nettles M. J. Zemanick K. Pioglitazone for Hepatic Steatosis in HIV/Hepatitis C Virus Coinfection AIDS Res. Hum. Retroviruses 2015 31 10 961 966 10.1089/aid.2015.0093 26214341 PMC4576938

Kamolvisit S. Chirnaksorn S. Nimitphong H. Sungkanuparph S. Pioglitazone for the Treatment of Metabolic-Associated Fatty Liver Disease in People Living With HIV and Prediabetes Cureus 2021 13 10 e19046 10.7759/cureus.19046 34858740 PMC8613454

Setz C. Große M. Auth J. Fröba M. Rauch P. Bausch A. Synergistic Antiviral Activity of Pamapimod and Pioglitazone against SARS-CoV-2 and Its Variants of Concern Int. J. Mol. Sci. 2022 23 12 6830 10.3390/ijms23126830 35743273 PMC9224751

Cefprozil: Uses, Interactions, Mechanism of Action|DrugBank Online . Available from: https://go.drugbank.com/drugs/DB01150 (cited Jun 27, 2025 ).

Cherkasov A. Muratov E. N. Fourches D. Varnek A. Baskin I. I. Cronin M. Dearden J. Gramatica P. Martin Y. C. Todeschini R. QSAR modeling: where have you been? Where are you going to? J. Med. Chem. 2014 57 12 4977 5010 10.1021/jm4004285 24351051 PMC4074254

Kitchen D. B. Decornez H. Furr J. R. Bajorath J. Docking and scoring in virtual screening for drug discovery: methods and applications Nat. Rev. Drug Discov. 2004 3 11 935 949 10.1038/nrd1549 15520816

Vilar , S. ; Costanzi , S. Predicting the biological activities through QSAR analysis and docking-based scoring . In Membrane Protein Structure and Dynamics: Methods and Protocols ; Humana Press : Totowa, NJ , 2012 ; pp 271 – 284 . 10.1007/978-1-62703-023-6_16 PMC3445294 22976034

Ancuceanu R. Tamba B. Stoicescu C. S. Dinu M. Use of QSAR global models and molecular docking for developing new inhibitors of c-src tyrosine kinase Int. J. Mol. Sci. 2020 21 1 19 10.3390/ijms21010019 PMC6981969 31861445

Filimonov D. A. Lagunin A. A. Gloriozova T. A. Rudik A. V. Druzhilovskii D. S. Pogodin P. V. Poroikov V. V. Prediction of the biological activity spectra of organic compounds using the PASS online web resource Chem. Heterocycl. Compd. 2014 50 3 444 457 10.1007/s10593-014-1496-1

Ko Y. Computational drug repositioning: current progress and challenges Appl. Sci. 2020 10 15 5076 10.3390/app10155076

Weth F. R. Hoggarth G. B. Weth A. F. Paterson E. White M. P. Tan S. T. Peng L. Gray C. Unlocking hidden potential: advancements, approaches, and obstacles in repurposing drugs for cancer therapy Br. J. Cancer 2024 130 5 703 715 10.1038/s41416-023-02502-9 38012383 PMC10912636

Guidotti I. L. Neis A. Martinez D. P. Seixas F. K. Machado K. Kremer F. S. Bambu and its applications in the discovery of active molecules against melanoma J. Mol. Graph. Model. 2023 124 108564 10.1016/j.jmgm.2023.108564 37453311

Trott O. Olson A. J. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading J. Comput. Chem. 2010 31 2 455 461 10.1002/jcc.21334 19499576 PMC3041641

Morris G. M. Goodsell D. S. Halliday R. S. Huey R. Hart W. E. Belew R. K. Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function J. Comput. Chem. 1998 19 14 1639 1662 10.1002/(SICI)1096-987X(19981115)19:14<1639::AID-JCC10>3.0.CO;2-B

Jakubec D. Skoda P. Krivak R. Novotny M. Hoksza D. PrankWeb 3: accelerated ligand-binding site predictions for experimental and modelled protein structures Nucleic Acids Res. 2022 50 W1 W593 W597 10.1093/nar/gkac389 35609995 PMC10353840

Adasme M. F. Linnemann K. L. Bolz S. N. Kaiser F. Salentin S. Haupt V. J. Schroeder M. PLIP 2021: expanding the scope of the protein–ligand interaction profiler to DNA and RNA Nucleic Acids Res. 2021 49 W1 W530 W534 10.1093/nar/gkab294 33950214 PMC8262720

Ivanova L. Karelson M. The Impact of Software Used and the Type of Target Protein on Molecular Docking Accuracy Molecules 2022 27 24 9041 10.3390/molecules27249041 36558174 PMC9788237

Zhu T. Cao S. Su P. C. Patel R. Shah D. Chokshi H. B. Hit Identification and Optimization in Virtual Screening: Practical Recommendations Based Upon a Critical Literature Analysis J. Med. Chem. 2013 56 17 6560 6572 10.1021/jm301916b 23688234 PMC3772997

Xiong G. Wu Z. Yi J. Fu L. Yang Z. Hsieh C. Yin M. Zeng X. Wu C. Lu A. ADMETlab 2.0: an integrated online platform for accurate and comprehensive predictions of ADMET properties Nucleic Acids Res. 2021 49 W1 W5 W14 10.1093/nar/gkab255 33893803 PMC8262709