AUTHOR=Ahmed Faheem , Lee Jae Wook , Samantasinghar Anupama , Kim Young Su , Kim Kyung Hwan , Kang In Suk , Memon Fida Hussain , Lim Jong Hwan , Choi Kyung Hyun 

TITLE=SperoPredictor: An Integrated Machine Learning and Molecular Docking-Based Drug Repurposing Framework With Use Case of COVID-19

JOURNAL=Frontiers in Public Health

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.902123

DOI=10.3389/fpubh.2022.902123

ISSN=2296-2565

ABSTRACT=The global spread of SARS Coronavirus 2 (SARS-CoV-2), its manifestation in human hosts as a contagious disease, and its variants have induced a global pandemic resulting in the deaths of over 6,000,000 people. Extensive efforts have been devoted to drug research to cure and refrain the spread of Covid-19, but only one drug has received FDA approval yet. Traditional drug discovery and development paradigms are inefficient, costly, and are not agile enough to react to pandemic threats. Drug repurposing represents an effective strategy of drug discovery and reduces the time and cost compared to de-novo drug discovery. In present study, a generic drug repurposing framework (SperoPredictor) has been developed which systematically integrates the various types of drugs and disease data and takes the advantage of machine learning (Random Forest, Tree Ensemble, and Gradient Boosted Trees) to repurpose potential drug candidates against any disease of interest. Drug and disease data for FDA approved drugs (n=2865) containing four drug features (Drug structures, drug target sequences, drug side effects, and drug related genes) and three disease features (Disease gene sequences, disease observable traits, and specificity index values) were collected from chemical and biological databases. Extracted data was integrated in the form of drug-disease association tables. Resulting dataset was split into 70% for training, 15% for testing, and remaining 15% for validation. The testing and validation accuracies of the models were 99.3% for Random Forest and 99.03% for Tree Ensemble. In practice, SperoPredictor identified 25 potential drug candidates against 6 human host target proteomes identified from a systematic review of peer reviewed journals. Literature-based validation indicated 12 of 25 predicted drugs (48%) have been already used for Covid-19 followed by molecular docking and redocking which indicated 4 of 13 drugs (30%) as a potential candidate against Covid-19 to be pre-clinically and clinically validated. Finally, SperoPredictor results illustrate the ability of the platform to be rapidly deployed to repurpose the drugs as rapid response to emergent situations (like Covid-19 and other pandemics).