Informatics Seminar: Computational Models to Identify Antibiotic Resistant Bacteria

Abstract: Antibiotic-resistant bacteria cause over 2 million infections and 23,000 deaths per year in the United States. The ability to identify antibiotic resistant bacteria based on mutations in their genome sequences is important for diagnosis, surveillance, and drug design. Current computational techniques often lack the statistical power to identify resistance-conferring mutations, especially when generalizing to newly evolved mutations. My research efforts focus on building the next generation of computational methods to predict antibiotic resistance, aiming to improve performance by integrating knowledge of biology into modern machine learning techniques. Currently, I am working to detect evolution towards new traits, especially resistance to new antibiotics, by combining multiple data modalities (genome sequence data and protein 3D-structural data). In the future, I plan to continue building methods that use biology-informed inductive biases to accurately and interpretably predict antibiotic resistance, including models that generalize across bacterial species, and natural language processing models of gene regulatory sequences.

Bio: Anna G. Green is a computational biologist who is passionate about building accurate and interpretable models to study the variation encoded in genomes. She holds a B.S. in Molecular and Cell Biology, with thesis research in bioinformatics, from the University of Connecticut. She completed her PhD in Systems Biology at Harvard University where she developed and applied computational approaches to predict the structure and interaction of proteins from biological sequence data. As a postdoctoral fellow at Harvard Medical School Department of Biomedical Informatics, she builds computational and machine learning approaches to study antibiotic-resistant bacteria from their genome sequences.