Big Data Predictive Phylogenetics with Bayesian Learning Abstract Andrew Holbrook, Ph.D., is a Bayesian statistician with a broad background in applied, theoretical and compu- tational data science. His proposed research Big Data Predictive Phylogenetics with Bayesian Learning tackles viral outbreak forecasting by combining Bayesian phylogenetic modeling with ?exible, `self-exciting' stochastic process models. The development and publication of open-source, high-performance computing software for his models will facilitate fast epidemiological ?eld response in a big data setting. Dr. Holbrook will apply his method- ology to the reconstruction of the 2015-2016 Zika virus epidemic in the Americas, focusing on identifying key geographical routes of transmission and phylogenetic clades with enhanced infectiousness. Candidate: Dr. Holbrook is Postdoctoral Scholar at the UCLA Department of Human Genetics. He earned his Ph.D. in Statistics from the Department of Statistics at UC Irvine, during which time he completed his dissertation Geometric Bayes, an investigation into Bayesian modeling and computing on abstract mathematical spaces, and simultaneously participated in scienti?c collaborations at the UC Irvine Alzheimer's Disease Research Center. The proposed career development plan will establish Dr. Holbrook as an independent leader in data intensive viral epidemiology by 1) facilitating coursework to build biological domain knowledge, 2) affording Dr. Holbrook the opportunity to lead his own project while remaining under the expert oversight of UCLA Prof. Marc Suchard, M.D., Ph.D., and 3) allowing Dr. Holbrook to continue his focus on quantitative viral epidemiology once he has moved to a faculty commitment. Mentors: During the ?rst three years of the award period, Dr. Holbrook will work closely with Prof. Suchard, continuing their current schedule of weekly meetings. Prof. Suchard is a leading expert in both Bayesian phylo- genetics and high-performance statistical computing; and with his medical background, Prof. Suchard will advise Dr. Holbrook in his expansion of domain knowledge in viral epidemiology. As secondary mentor, Prof. Kristian Andersen, Ph.D., of the Scripps Institute will advise Dr. Holbrook in the impactful application of his statistical and computational methodologies to the 2015-2016 Zika virus epidemic. Dr. Holbrook and Profs. Suchard and Andersen will maintain their collaborations after the postdoctoral period. Research: Bayesian phylogenetics successfully reconstructs evolutionary histories but fails to predict viral spread. Self-exciting point processes are devoid of biological insight and fail to account for geographic networks of diffusion.
Aim 1 addresses de?ciencies in these two complementary viral epidemiological modeling techniques by innovating a combined model where the phylogenetic and self-excitatory components support each other.
Aim 2 makes widespread adoption a reality by publishing open-source, massively parallel computing software suitable for big data analysis.
Aim 3 reconstructs the 2015-2016 Zika epidemic, learns key geographical routes of transmission and identi?es phylogenetic clades with enhanced infectiousness.
Tracking and predicting viral outbreaks remains an open epidemiological problem with deadly consequences. Dr. Holbrook will attack the problem with his Bayesian phylogenetic Hawkes processes, a class of models tailored to simultaneously reconstruct evolutionary histories and predict viral diffusion dynamics. With the mentorship of Profs. Marc Suchard (primary) and Kristian Andersen (secondary), Dr. Holbrook will develop open-source, high-performance computing software and apply his statistical computing methodology to the analysis of the 2015-2016 Zika virus epidemic of the Americas, learning key routes of transmission and identifying phylogenetic clades with enhanced infectiousness.
