RNA-seq analysis of knock-out model

RNA-seq libraries were prepared using 200 ng of total RNA as described in Nikolaeva et al.^{5 (link)}. Illumina TruSeq SR Cluster Kit v4 reagents were used. Sequencing data were processed as described in^{45 (link)} using Mus musculus.GRCm38.86 gene annotation. Statistical analysis was performed in R (version 3.4.0). Genes with low counts were filtered out according to the rule of 1 count per million (cpm) in at least 1 sample. Library sizes were scaled using TMM normalization and log-transformed into counts per million (CPM) using voom^{46 (link)}.
Principal component analysis showed a large variability between replicate samples. Two factors of unwanted variation were removed using the RUVs function (R package RUVSeq v. 1.10.0^{47 (link)}). Differential expression between knock-out and wild-type was computed using limma^{48 (link)}. A linear model with a factor for each combination of time point and genotyope was used. Factors to correct for batch effect and unwanted variation were also added in the design matrix. Differences in gene expression levels between KO and WT animals at each time points were combined into one F-test. Genes with a false discovery rate <5% were considered significant. The limma function ‘classifyTestsF’ was used to classify time point as significant or not for the selected genes.