Identifying GCDEGs using SeqGene on RNA-Seq data. (A). Comparing variance components of GCDEGs (white bars) with DEGs (dark bars). The variance components were computed for three factors (Treatment, Genotype, and Residue) using SeqGene with the 'lme' function in R package 'nlme' and the 'varcomp' function in R package 'ape'. In this example, the uncontrolled DEGs were detected using Student's t-test p < 0.01. The controlled GCDEGs satisfied both Student's t-test p < 0.01 and the adjusted p-value < 0.01. We observed that Residue variance is significantly reduced in GCDEGs. (B). An example gene showing the effect of avoiding the confounding genotype factor. In this example, the treatment and genotype are completely confounded in that the treatment samples had genotype (C/T) and the control samples had genotype (T/T). The uncontrolled p-value is 0.0015, whereas it is no longer significant after genotype controlling.
Deng BMC Bioinformatics 2011 12:267 doi:10.1186/1471-2105-12-267