In conjunction with the interdisciplinary activities of the National Institute for Mathematical and Biological Synthesis (NIMBioS), a seminar series on topics in mathematical biology will be hosted at NIMBioS every other Tuesday at 3:30 p.m. (unless otherwise noted) in the Hallam Auditorium, Room 206, Claxton Building, 1122 Volunteer Blvd. Seminar speakers will focus on their research initiatives at the interface of mathematics and many areas of the life sciences. Light refreshments will be served in Room 205 beginning 30 minutes before each talk. Faculty and students from across the UT community are welcome to join us.
Speaker: Dr. Scott J. Emrich , Bioinformatics, Univ. of Notre Dame
Time/Date: Tuesday, October 17, 2017, 3:30*
Location: Room 206, Claxton Building, 1122 Volunteer Blvd.
Topic: Integrating diverse data for improved computational genomics
Abstract: Genomics-driven analysis of many important species, which we have called "non-models", remains challenging. My group is funded by the NIH to computationally leverage and enhance newer higher-throughput sequencing with domain expert-provided metadata including biological phenotypes (e.g., drug resistance in malaria), protein folding experiments, and karyotype data from mosquitoes.
For this talk I will focus on updates to our 2016 ACM BCB paper, which has been recently published in IEEE/ACM Transactions on Computational Biology and Bioinformatics (TCBB). Previous computational approaches for imputation of missing genotype data have relied on a linear order of markers and a genotype panel, both of which are not common in non-models. We address this limitation with our ADDIT (Accurate Data-Driven Imputation Technique) approach, which is composed of two data integration-focused algorithms: a non-model variant that employs statistical inference, and a model organism variant that better leverages reference data using a supervised learning-based approach. I will show that ADDIT is more accurate, faster and requires less memory than state-of-the art methods using model (human) and non-model (maize, apple, grape) datasets. I will conclude with emerging "-omics" results from a few other current projects, most involving either a mosquito vector or the disease they transmit. The methods we are developing combine traditional sequence-based analysis with existing and new models, mostly network-based, through collaborations to best integrate diverse data types. I will discuss some early successes and challenges, with the latter hopefully a start for new discussions in Knoxville come January.
Prior to joining the faculty of the University of Tennessee (Knoxville), Scott Emrich was the Director of Bioinformatics at University of Notre Dame with a primary appointment in Computer Science and Engineering and a courtesy appointment in Biology. He obtained a BS in Biology and Computer Science from Loyola College (MD) and a 2007 PhD in Bioinformatics and Computational Biology from Iowa State University. He has published over 78 peer-reviewed publications including venues such as "Science" (4, 2 covers), "PNAS" (3), "Nature and Genome Research". Most of his genomics-focused work at Notre Dame involved his twelve PhD students (9 grads) in a leadership role. In fact, two of his students co-won interdisciplinary submissions for both the Assemblethon2 and the DREAM regulatory prediction challenge. He has four active awards from the NIH.
*Join us for refreshments at 3 p.m.
Seminar Flyer (pdf)
For more information about this and other NIMBioS Seminars, visit http://www.nimbios.org/seminars.