Time Course Models for RNA Sequencing Data

• Activity Number: 629
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Andrew Jaffe*+
• Companies: Lieber Institute for Brain Development
• Keywords: gene expression ; RNA sequencing ; time course ; spatial correlation ; multiple testing ; computational biology

Abstract

Recent advances in sequencing technologies now allow for the complete and quantitative characterization of the transcriptional landscape of biological samples. This technology, RNA sequencing (RNAseq), produces large datasets (tens of millions of observations per sample) potentially affected by technical artifacts that are not present in previous microarray gene expression data. Here we develop statistical models specifically for "time course" biological experiments, where multiple measurements are taken from replicates across time to better isolate particular cellular exposures or trajectories. We focus on two popular designs: 1) "single sample", where one subject is serially sampled over many intervals and 2) "developmental", where pluripotent cells in a dish are differentiated to various outcomes using precise treatment regimes, and there are often several biological replicates each at fewer time points. The large quantity of data, and its intrinsic spatial correlation across both the genome and time demonstrate the need for these special statistical models, that both properly control for multiple testing and also validate the known biology of the data under study.