Abstract: Metabolomics is a powerful approach to characterize small molecules produced in cells, tissues, and other biological systems. Metabolites are direct products of enzymatic reactions and provide a snapshot of cellular activities. Metabolomics-based research has already had a profound impact on biomarker discovery, nutritional analysis, and other biomedical and biological discoveries. The most pressing problem in metabolomics however is identifying compounds in the sample-under-study from the metabolomics measurements. Current analysis tools are capable of annotating only a small portion of sample measurements.

In this talk, we present machine learning solutions to three challenges related to the interpretation of metabolomics data. To mimic the function of a mass spectrometer in generating a mass spectrum, we use graph neural networks to translate a molecular structure into its respective spectral signature. To interpret the biological measurements in the context of the biological sample, we use Bayesan learning to deduce the likelihood of pathway activities. To suggest putative candidate molecules that are biologically relevant matches to the measured spectra, we explore several methods for predicting possible enzymatic products. We discuss several results, highlighting the value of using machine learning for advancing metabolomics analysis.

Short bio: Soha Hassoun is Professor and Past Chair of the Department of Computer Science at Tufts University. Soha received her undergraduate degree in Electrical Engineering from South Dakota State University, the Master's degree from MIT, and the Ph.D. degree from the Department of Computer Science and Engineering, University of Washington in Seattle. Soha’s lab uses Machine Learning to develop analysis and discovery tools for synthetic and systems biology, with a focus on enzyme promiscuity prediction and metabolomics analysis. Soha was a recipient of the NSF CAREER Award, and several technical and service awards from various professional societies. She provided technical leadership for several conferences including ICCAD and DAC. She co-founded the International Workshop on Bio-Design Automation in 2009. Soha serves on the board of the Computing Research Association's Committee on Widening Participation in Computing Research.

https://umich-health.zoom.us/j/93929606089?pwd=SHh6R1FOQm8xMThRemdxTEFMWWpVdz09