how does life work?
Living systems are complex, spanning vast distances in space and time, and integrating processes that scale from subatomic to global. Making sense of biological systems like cells, organisms, communities, ecologies and evolutionary histories is challenging. Most of my time is spent thinking about how to incorporate existing knowledge about how biological systems work into new inference paradigms like artificial intelligence, in order to learn new things about how biological systems work.
I see science as a creative - as well as technical - discipline. It all starts with an idea, and the best ideas challenge dominant paradigms in fresh and interesting ways. Of course, most ideas turn out to be wrong - or at least incomplete - when rigorously tested. But the rewards of doing science come from constantly pushing yourself into the unknown to make new discoveries.
Bryan finished his PhD at the University of Oregon in 2006 under an NSF fellowship in Evolution, Development and Genomics that cobbled together ideas from computer science and molecular evolution to create one of the first research degrees in 'bioinformatics'. After completing postdoctoral research at UC Berkeley and Dartmouth College, he ended up running a lab at University of Florida as part of a campus-wide effort to increase research and education in 'computational biology'. Bryan’s work remains at the forefront of scientific buzzwords, as he continues to help lead UF’s efforts to build research and educational infrastructure in 'artificial intelligence'.
I'm good at:
here are some of my recent projects
Antiviral Immune Receptors Remain Open to Change
The evolutionary ‘arms race’ between pathogens and their hosts has left indellible marks on our genomes that impact how our bodies respond to new pathogens. By combining ancestral sequence reconstruction, structural modeling and dynamics, and machine learning, our lab was able to shed new light on how our immune system evolved to reliably identify existing pathogens while maintaining the ability to respond to new threats. We found that key antiviral immune receptors maintain functional plasticity by repeatedly ‘flip-flopping’ among a small number of structural-functional states over short evolutionary timescales. Better understanding the structural and functional evolution of immune receptors is expected to help us better understand, predict and respond to pathogen outbreaks.
Fake Data Helps Neural Networks Predict Real Disease
Predicting disease risk from personalized genomic and metagenomic data has the potential to revolutionize medicine. Deep neural networks can provide amazingly accurate statistical inferences but require way more training data than we have. We are developing generative adversarial networks to model the distributions of biomedical data associated with disease risk, allowing us to generate essentially unlimited amounts of training data for disease-risk prediction. Using deep AI to better understand disease risk is expected to help us unravel the complex relationships among our genomes, our metagenomes and how our bodies function. For more information about this project, check out this short video.