Abstract
The inhibition of protein–protein interactions (PPIs) continues to be a major area of investigation given the importance of these systems in both normal and adverse biological processes. Interest in this area continues to grow rapidly and perceptions regarding the feasibility and success of PPI modulation have changed given high-profile clinical successes with drugs like navitoclax and venetoclax. Interestingly, these PPI success stories are based on the principle that most of the free energy of the binding interactions has been mediated by binding “hotspots” (HSs) that can be targeted with small molecules. Historically, identifying PPI HSs and evaluating their druggability have been carried out at the bench, one PPI pair at a time, over a period of months or years. While computational methods such as protein–protein docking have been developed to assist, there is currently a major gap in the knowledge around PPIs since the current approaches depend on three-dimensional structural data and often suggest hundreds of potential PPI interfaces. To address current limitations in locating druggable PPI interfaces, we propose a novel platform that uses experimentally determined HSs and nullspots (NSs) considering 51 crystallized protein complexes. With these data, we have built a robust classifier using machine learning (ML) approaches to predict new HSs, followed by molecular dynamics simulations (Site-Map) to identify tractable small molecule binding pockets.
To validate our ML-based classifier we considered the PPI sites between the mutant p53 and its stabilizing chaperone DNAJA1 and identified the HSs. Then the HS sites are considered to define the small molecule ligand binding site to screen the curated drug-like library and identified the candidates for pancreatic cancer from our in-vivo experiments.
Original language | English (US) |
---|---|
Title of host publication | Big Data Analytics in Chemoinformatics and Bioinformatics |
Subtitle of host publication | with Applications to Computer-Aided Drug Design, Cancer Biology, Emerging Pathogens and Computational Toxicology |
Publisher | Elsevier |
Pages | 247-263 |
Number of pages | 17 |
ISBN (Electronic) | 9780323857130 |
ISBN (Print) | 9780323857147 |
DOIs | |
State | Published - Jan 1 2022 |
Keywords
- Big data
- drug-like molecule
- homology modeling
- hot-null spots
- machine learning
- pancreatic cancer
- protein–protein docking (PPD)
- virtual high throughput screening (vHTS)
ASJC Scopus subject areas
- General Chemistry