Wednesday, June 22, 2016
Dr. Fredrik Vannberg (School of Biology)
Dr. King Jordan (School of Biology)
Dr. Srinivas Aluru (School of Computational Science and Engineering)
Dr. Brian Hammer (School of Biology)
Dr. Greg Gibson (School of Biology)
Modern genomic experiments are taking advantage of rapid growth in sequencing technology throughput. With so much data generated by a single sequencing run, many scientists are turning to distributed and cloud computing platforms to process it. Instead of finding new ways to dedicate more resources to this problem, Peter Audano will discuss new algorithms that reduce the computational burden of analysis. Most approaches begin by aligning sequence reads to a reference, but with an emerging class of mapping-free algorithms, hours of CPU time can be saved by skipping this step. This defense will discuss the fundamental k-mer data structures that support these methods and new variant calling techniques that work efficiently even where alignments fail.