Prediction of repurposed drugs for Coronaviruses using artificial intelligence and machine learning

Area of Science:

• Computational chemistry and bioinformatics
• Machine learning in drug discovery
• Virology and infectious diseases

Background:

• The Coronaviridae family, including SARS-CoV-2, poses significant global health threats.
• There is an urgent need for effective antiviral therapies against coronaviruses.
• Existing antiviral drug options for these viruses are limited.

Purpose of the Study:

• To develop and validate machine learning models for predicting repurposed drugs against coronaviruses.
• To identify novel drug candidates for treating infections caused by SARS-CoV-2, SARS, and MERS.
• To accelerate antiviral drug discovery through computational approaches.

Main Methods:

• Utilized machine learning algorithms (SVM, Random Forest, k-NN, ANN, Deep Learning) for drug repurposing prediction.
• Extracted and filtered chemical descriptors and fingerprints from experimentally validated anti-coronavirus compounds.
• Validated predictive models using independent datasets, decoy sets, and molecular docking against the SARS-CoV-2 spike protein-ACE receptor complex.

Main Results:

• Developed robust predictive models with Pearson's correlation coefficients ranging from 0.60 to 0.90.
• Identified several promising repurposed drug candidates, including Verteporfin, Alatrofloxacin, Metergoline, Rescinnamine, Leuprolide, and Telotristat ethyl.
• These candidates demonstrated high binding affinity to the SARS-CoV-2 spike protein.

Conclusions:

• The developed computational models are effective tools for identifying potential antiviral drug candidates.
• Repurposed drugs like Verteporfin show promise for treating SARS-CoV-2 and other coronavirus infections.
• This approach can significantly aid in the rapid discovery of new antiviral therapies.