RNA-sequence data normalization through in silico prediction of reference genes: the bacterial response to DNA damage as case study.

摘要: Measuring how gene expression changes in the course of an experiment assesses organism responds on a molecular level. Sequencing RNA molecules, and their subsequent quantification, aims to assess global level (transcriptome). While advances high-throughput RNA-sequencing (RNA-seq) technologies allow for inexpensive data generation, accurate post-processing normalization across samples is required eliminate any systematic noise introduced by biochemical and/or technical processes. Existing methods thus either normalize selected known reference genes that are invariant experiment, assume majority invariant, or effects up- down-regulated cancel each other out during normalization. Here, we present novel method, moose 2 , which predicts silico through dynamic programming (DP) scheme applies quadratic based this subset. The method allows specifying set experimentally validated genes, guides DP. We verified predictions bacterium Escherichia coli, show able (i) estimate value distances between RNA-seq samples, (ii) reduce variation values all (iii) subsequently reveal new functional groups late stages DNA damage. further applied three eukaryotic sets, its performance compares favourably methods. software implemented C++ publicly available from http://grabherr.github.io/moose2/ . proposed valuable alternative existing methods, with two major advantages: prediction provides list potential downstream analyses, non-linear artefacts handled adequately minimize variations replicates.