Higher order asymptotics for negative binomial regression inferences from RNA-sequencing data.

摘要: RNA sequencing (RNA-Seq) is the current method of choice for characterizing transcriptomes and quantifying gene expression changes. This next generation sequencing-based provides unprec- edented depth resolution. The negative binomial (NB) probability distribution has been shown to be a useful model frequencies mapped RNA-Seq reads consequently basis statistical anal- ysis expression. Negative exact tests are available two-group comparisons but do not extend regression analysis, which important examining as func- tion explanatory variables adjusted group accounting other factors. We address adequacy large-sample small sample sizes typically from studies consider higher-order asymptotic (HOA) adjustment likelihood ratio tests. demonstrate that 1) HOA-adjusted test practically indistinguishable in situations where available, 2) type I error HOA matches nominal specification settings we examined via simulation, 3) power does appear affected by adjustment. work helps clarify accuracy unadjusted degree improvement with Furthermore, may preferable even when because it require ad hoc library size adjustments.