Treatment options for people infected with antibiotic-resistant bacteria today are limited, and may become even more so over time. The risk of a post-antibiotic era, in which even minor infections or routine medical procedures could be fatal, is real. Beyond the personal costs, the economic burden of drug resistance is high, estimated to cost more than $1 billion each year in North America alone.
Genomic screening of pathogens to determine their identity and responsiveness to antibiotics could be both more accurate and efficient than other ways of diagnosing infectious disease and choosing the appropriate treatment. The challenge, though, lies in two areas: analyzing the large amount of data produced by this sequencing to understand the genomic factors underlying drug resistance, and creating an accurate model to predict drug resistance based on sequencing data.
Drs. Leonid Chindelevitch and Maxwell Libbrecht of Simon Fraser University and Jesse Shapiro of University of Montreal are developing computational tools based on machine learning, which will be able to unravel the complex relationships between bacterial genome sequences and antibiotic resistance, based on bacterial genome data currently available in public databases as well as from their collaborators.
The project, which builds on two projects funded in the 2015 bioinformatics and computational biology competition, will result in two tools: a comprehensive predictive model for resistance to a variety of drugs for specific bacteria, which will update itself as new data become available; and a user-friendly web interface to enable researchers, including clinical and public health researchers, to securely upload and analyze genomic data from pathogenic bacteria and accurately predict drug resistance. The tool looks to boost research in drug resistance and development of better clinical tools to the ultimate benefit of infectious disease patients.