Description:
- Drug discovery
- Antibacterial resistance
- Treatment of life-threatening and rare diseases
Abstract
USC researchers have developed a machine learning approach that can infer drug properties from known information about drug-target interactions. The technique is used to build a weighted Drug-Drug Similarity Network that can generate drug communities associated with specific, dominant drug properties. Drugs assigned to a community which don’t match the community’s dominant pharmacologic property may indicate the need for repurposing. This method was used to identify the anxiolytic Meprobamate as a possible antifungal, a new application supported by results of a molecular docking test.
Benefit
- Predicts new applications of existing drugs
- Reduces costs associated with drug development
- Reduces medication risks
- Prepares for wet lab experiments
Market Application
While the number of new FDA drug applications has surged in the last decade, the number of approved drugs has grown only marginally. This expensive and time-consuming disconnect indicates the need to optimize the drug development process. One possible avenue is finding new pharmaceutical applications for already approved drugs, a process called drug repositioning or drug repurposing. Computational methods have emerged as a powerful tool for drug repositioning.
Publications
Uncovering New Drug Properties in Target-Based Drug–Drug Similarity Networks, Udrescu et al., 2020.
Other
Stage of Development
- Software developed
- Results validated in silico
- Available for licensing