Mitigating the effects of climate change on public health and conservation calls for a better understanding of the dynamic interplay between biological processes and environmental effects. The state-of-the-art, which has led to many important discoveries, utilizes numerical or statistical models for making predictions or performing in silico experimentation, but these techniques struggle to capture the nonlinear response of natural systems. Machine learning (ML) methods are better able to cope with nonlinearity and have been used successfully in biological applications, but several barriers still exist, including the opaque nature of the algorithm output and the absence of ML-ready data. This project seeks to significantly advance technologies in ML and create a new interdisciplinary field, computational ecogenomics. This will be accomplished by designing ML techniques for encoding heterogeneous genomic and environmental data and mapping them to multi-level phenotypic traits, reducing the amount of necessary training data, and then developing interactive visualizations to better interpret ML models and their outputs. These advances will responsibly and transparently inform policy to maximize resources during this crucial window for planetary health, while revealing underlying biological mechanisms of response to stress and evolutionary pressure.
The long-term vision for this project is to develop predictive analytics for organismal response to environmental perturbations using innovative data science approaches and change the way scientists think about gene expression and the environment. The goal for this two-year award is to develop a proof-of-concept for an institute focused on predicting emergent properties of complex systems; an institute that would itself foster the development of many new sub-disciplines. The core of this activity is developing a machine learning framework capable of predicting phenotypes based on multi-scale data about genes and environments. Available data, ranging from simple vectors to complex images to sequences, will be ingested into this framework by applying proven semantic data integration tools and algorithmic data transformation methods. The central hypothesis of this research is that deep learning algorithms and biological knowledge graphs will predict phenotypes more accurately across more taxa and more ecosystems than do current numerical and traditional statistical modeling methods. The rationale for this project is that a timely investment in data science will push through a bottleneck in life science, accelerating discovery of gene-phenotype-environment relationships, and catalyzing a new computational discipline to uncover the complex "rules of life."
This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity, and is jointly supported by the HDR and the Division of Biological Infrastructure within the NSF Directorate of Directorate for Biological Sciences.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
