Skip to main content
Figure 1 | BMC Genomics

Figure 1

From: Putative null distributions corresponding to tests of differential expression in the Golden Spike dataset are intensity dependent

Figure 1

A two-stage procedure was used to obtain the sets of null p-values referenced in [1] and [2]. The first stage of the procedure involves the application of algorithms designed to correct and normalize the raw data matrix, W. The second stage of the procedure involves the evaluation of a test statistic for differential expression using information from the Stage II data matrix, X. [1] considered 150+ unique combinations of algorithms and input parameter values for the Stage I analysis and proposed a subset of the 10 best. [2] determined that the distributions of the null p-values for the [1] 10 best Stage I analyses were non-uniform for the most common choices for the Stage II analysis (e.g., t-test, permutation t-test, and Wilcoxon Rank Sum test). [2] concluded that this non-uniformity implies that the technical replicates generated by the experiment do not constitute adequate approximations of biological replicates and hence, the Stage II model assumptions associated with these tests are not met. We provide evidence that the non-uniformity can be attributed, at least in part, to the failure of the Stage I analyses to correct for systematic biases in the raw data matrix, W.

Back to article page