Computer Science PhD student Daniel Knights was selected as the winner of the College of Engineering and Applied Science's Outstanding Dissertation Award.
The award was given at the May 2012 Engineering Recognition Ceremony. Dan's thesis, entitled "Predictive Modeling of Metagenomes", involves the use of machine learning and statistical techniques to construct models of complex microbial communities. These models facilitate the understanding and classification of high-throughput sequencing data of multiple microbial communities, and have applications in personalized medicine, early disease diagnosis, and biofuel production.
Dan basically introduced machine learning into a field that needed but lacked it. Use of these powerful techniques allowed him to reveal, for instance, that an individual's gut microbes allow one to predict with 90% accuracy whether the individual is lean or obese, and the skin microbes allow one to predict whether the individual will be attractive to mosquitos. His algorithm for Bayesian community-level microbial source tracking has attracted interest from the EPA, USDA, CDC, Department of Justice, the Army and Navy Sponsored Research Offices, as well as several other agencies that don't let their employees identify their employer. Many of his techniques are now a standard part of the microbial ecology workflow, and are being used to address problems ranging from microbes associated with persistent malnutrition in children in Malawi and Bangladesh, to global climate change, to microbial fuel cells.
Dan is a co-author on sixteen journal publications that have appeared in 2010 and 2011, including two Science papers, one Proceedings of the National Academcy of Sciences paper, and all in top tier journals (e.g.,Nature Methods). Dan is first author on four of these publications, and in almost every one of his sixteen publications, he is the primary computational modeler involved in the research. Dan has also played a key role in the Human Microbiome Project, a $175-million NIH initiative, performing a substantial amount of the analysis for the primary papers currently in revision at Nature. His HMP analyses were chosen for a talk at the International Human Microbiome Congress in Paris this March, and highlighted in a news item in Nature. His PNAS paper on bioreactor communities was labeled a "must read" by the Faculty of 1000. His paper on mosquito attractiveness (PLoS ONE) received media coverage from NPR, Scientific American, Discover Magazine, and MSNBC, and is listed first on the card that PLoS ONE editors hand out to promote their microbiome publications. His work on microbial biogeography of public restroom surfaces was featured by NPR, Scientific American, and WebMD. It also prompted a call from the CU Physical Plant who expressed concern about the cleanliness of the two campus restrooms that he studied. His Science paper with Wu et al. was named by the journal as a runner-up for the publication's Breakthrough of the Year. Several of his other publications have been covered in Science News and Science Daily. Dan has given invited talks at universities across the US, and conferences around the world.
Dan graduated in the May ceremonies with a PhD in Computer Science and a Certificate in Interdisciplinary Quantitative Biology, the first of the IQ Bio students to graduate. Dan was co-advised by Professors Rob Knight (Chemistry & Biochemistry Department, Computer Science Department, and Biofrontiers Institute) and Michael Mozer (Computer Science and Institute of Cognitive Science).