Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Practical impacts of genomic data “cleaning” on biological discovery using surrogate variable analysis

Fig. 2

SVA improves power to identify differentially expressed genes. a PAX6 shows significant differential expression between mesendodermal and neurectodermal differentiation before SVA (p = 1.77 × 10−33) and b this effect becomes more significant following SVA (p = 4.82 × 10−53). c Prior to SVA, OLFML1 is not identified as being differentially expressed between differentiation conditions (p = 8.23 × 10−5), but d is highly significant after properly controlling for unwanted latent heterogeneity with SVA (p = 4.49 × 10−29). Expression values are on the log2 scale. Statistical significance was derived from a moderated t-statistic comparing expression in the mesendodermal differentiation condition versus that in the neurectodermal differentiation while also allowing variability to be explained by the undifferentiated condition (e.g. condition was categorical with 3 groups). Individual cell lines are represented on the X-axis. Gene expression on the Y-axis is depicted in quantile normalized, log2-scale intensities

Back to article page