Anti-Dengue: A Machine Learning-Assisted Prediction of Small Molecule Antivirals against Dengue Virus and Implications in Drug Repurposing

• Published in: Viruses Vol. 16; no. 1; p. 45
• Main Authors: Gautam, Sakshi, Thakur, Anamika, Rajput, Akanksha, Kumar, Manoj
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 27.12.2023; MDPI
• Subjects: Algorithms; Antiviral agents; Antiviral Agents - pharmacology; Artificial intelligence; Care and treatment; Chemicals; Datasets; Dengue; Dengue fever; Dengue Virus; Drug delivery; Drug development; Drug discovery; Drug Repositioning; Drugs; Ebola virus; Epidemics; Feature selection; Learning algorithms; Machine Learning; Prediction models; predictive models; QSAR; Regression analysis; Reproducibility of Results; Structure-activity relationships; Support vector machines; Tropical environment; Viruses; web server; India; predictive models; machine learning; QSAR; dengue virus; web server
• ISSN: 1999-4915; 1999-4915
• DOI: 10.3390/v16010045

Dengue outbreaks persist in global tropical regions, lacking approved antivirals, necessitating critical therapeutic development against the virus. In this context, we developed the "Anti-Dengue" algorithm that predicts dengue virus inhibitors using a quantitative structure-activity relationship (QSAR) and MLTs. Using the "DrugRepV" database, we extracted chemicals (small molecules) and repurposed drugs targeting the dengue virus with their corresponding IC values. Then, molecular descriptors and fingerprints were computed for these molecules using PaDEL software. Further, these molecules were split into training/testing and independent validation datasets. We developed regression-based predictive models employing 10-fold cross-validation using a variety of machine learning approaches, including SVM, ANN, kNN, and RF. The best predictive model yielded a of 0.71 on the training/testing dataset and 0.81 on the independent validation dataset. The created model's reliability and robustness were assessed using William's plot, scatter plot, decoy set, and chemical clustering analyses. Predictive models were utilized to identify possible drug candidates that could be repurposed. We identified goserelin, gonadorelin, and nafarelin as potential repurposed drugs with high pIC50 values. "Anti-Dengue" may be beneficial in accelerating antiviral drug development against the dengue virus.