Arun Ross awarded NSF grant
Arun Ross, Professor in the Department of Computer Science and Engineering, has been awarded a grant from the National Science Foundation titled "Collaborative Research: Converging Genomics, Phenomics, and Environments Using Interpretable Machine Learning Models”. Ross will be collaborating with Anne Thessen from Oregon State University (Lead PI), Patrick Heidorn from University of Arizona, and Remco Chang from Tufts University to develop machine learning methods that can predict the phenotypic attributes of an organism based on its genetic constitution and environmental conditions. The proposal was developed under NSF’s Harnessing the Data Revolution (HDR) program (I-DIRSE-IL).
The ways in which living things grow and behave are determined by a combination of their genes and the environment in which they live. These features and behaviors are called “phenotypes". A key scientific challenge is in predicting an organism's responses and resulting phenotypes to environmental perturbations, such as a flood or drought. Current phenotype prediction methods use statistics, but struggle to capture the chaotic complexity of natural systems and frequently do not include an organism's genetic information. Machine learning (ML) methods are better able to cope with complex biological systems and can incorporate genetic information, but require a lot of work to obtain and prepare the data; further, the prediction models themselves can be hard to interpret. This project will solve these problems by harnessing cyberinfrastructure-enabled computing and data storage, designing faster techniques for preparing biological data for ML, and developing interactive visualizations to better understand the prediction models and their outputs. If successful, these integrative advances will immediately, responsibly, and transparently inform policy to maximize resources. For example, American seed and livestock companies will be able to increase their competitiveness and natural resource managers will be able to better predict population fluctuations using this technology. Successful outcomes would also lead to the launch of an institute focused on making predictions about a much wider diversity of complex environmental-genetic systems.
(Date Posted: 2019-09-19)