APPLICATION OF MACHINE LEARNING APPROACHES TO EMPOWER DRUG DEVELOPMENT

Human health, one of the major topics in Life Science, is facing intensified challenges, including cancer, pandemic outbreaks, and antimicrobial resistance. Thus, new medicines with unique advantages, including peptide-based vaccines and permeable small molecule antimicrobials, are in urgent need. However, the drug development process is long, complex, and risky with no guarantee of success. Also, the improvements in techniques applied in genomics, proteomics, computational biology, and clinical trials significantly increase the data complexity and volume, which imposes higher requirements on the drug development pipeline. In recent years, machine learning (ML) methods were employed to support drug development in various aspects and were shown to be highly effective. Here, we explored the application of advanced ML approaches to empower the development of peptide-based vaccines and permeable antimicrobials. First, the peptide-based vaccines targeting pancreatic cancer and COVID-19 were predicted and screened via multiple approaches. Next, novel structure-based methods to improve the performance of peptide: MHC binding affinity prediction were developed, including an HLA modeling pipeline that provides structures for docking-based peptide binder validation, and hierarchical clustering of HLA I into supertypes and subtypes that have similar peptide binding specificity. Finally, the physicochemical properties governing the permeability of small molecules into multidrug-resistant Pseudomonas aeruginosa cells were selected using a random forest model. In conclusion, the use of machine learning methods could accelerate the drug development process at a lower cost and promote data-based decision-making if used properly.