|
Abstract:
|
Analyzing dose effects on gene expression plays a very important role in genomic dose-response analysis. The first challenge lies in identifying genes whose expression levels vary according to dose levels in a non-random manner. This is done in order to avoid dose response modeling of data with no statistically plausible signal. Traditionally the ANOVA test or a trend test is performed for bulk RNA-seq experiments. However, differences in gene expression between individual cells, are often lost in the averages of bulk RNA sequencing. Single-nuclei exploration of genomic dose-response analysis is therefore essential to interpret the causes and consequences of this cellular heterogeneity. Single nuclei data, however, is highly heterogeneous and has a large number of zero counts, which introduces new challenges in identifying dose-dependent differentially expressed genes. We therefore propose Bayesian extensions of the traditional ANOVA and trend tests to account for the dropouts and cell level heterogeneity exhibited by single nuclei RNA-sequencing datasets. Results from simulation studies and real experimental data shows superior performance over traditional testing methods.
|
