Computational cleaning of noisy 5' end tag sequencing data sets from rare in vivo cells

Abstract

We present a data filtration and processing pipeline for analysis of nanoCAGE-seq, a variant of the CAGE method allowing for very small amounts of input material, and thus expanding the number of tissue- and cell types available for proteome profiling. For this inherently noisy method, rigorous filter methods, including tag clustering, cluster width and profile filtering, and variance filtering rescue bona fide promoters, allowing for detection of promoter usage, inter-sample promoter switching and detection of new putative promoters. We present result from nanoCAGE from two different studies.