Novel Development of Predictive Feature Fingerprints to Identify Chemistry-Based Features for the Effective Drug...

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and development
• Infectious disease research, specifically COVID-19

Background:

• The COVID-19 pandemic necessitates rapid identification of effective therapeutics.
• Predictive modeling requires high-quality curated bioactivity and chemical data.
• Existing methods may lack cross-validation and target specificity for complex diseases.

Purpose of the Study:

• To develop and validate target-specific predictive feature fingerprints (PFF) for severe acute respiratory syndrome due to coronavirus 2 (SARS-CoV-2).
• To assess the predictability of these PFFs across multiple therapeutic targets and disease stages.
• To explore the utility of the developed models for virtual screening and drug repurposing.

Main Methods:

• Unique approach combining bioactivity and chemical data curation.
• Application of random forest analyses to generate predictive feature fingerprints (PFF).
• Cross-validation of PFFs against multiple therapeutic targets including plasma kallikrein, human immunodeficiency virus (HIV)-protease, nonstructural protein (NSP)5, NSP12, Janus kinase (JAK) family, and AT-1.

Main Results:

• Generated a series of target-specific and cross-validated PFFs with high predictability.
• Demonstrated high accuracy in matching compounds to their respective therapeutic targets.
• Successfully applied the curation-modeling process to a SARS-CoV-2 phenotypic screen.

Conclusions:

• The developed PFFs offer a robust tool for predicting compound activity against SARS-CoV-2 targets.
• The models are suitable for virtual screening of compound libraries and drug repurposing initiatives.
• This approach can guide clinical trial selection and the analysis of proprietary datasets.