Larry Ruzzo, University of Washington

The biological sciences have been/are being revolutionized by high-throughput, quantative measurement technologies. RNA sequencing is a poster-child example. RNAseq data is now widely used for analysis of gene expression, and is widely viewed as simple and quantitatively accurate. “You just count reads. What could go wrong?” A closer look at RNAseq data, however, quickly reveals *extensive* technical bias of unknown origin and consequence, as well as difficult problems inferring, e.g., gene isoform (alternative splicing) levels from read counts. My talk will outline a purely computational approach for quantifying and (partially) correcting for sequence-dependent biases and for robustly estimating isoform changes in a set of RNAseq measurements.