DESeq2 offers tests for composite null hypotheses of the form |βir|≤θ, where βir is the shrunken LFC from the estimation procedure described above. (See Materials and methods for details.) Figure 4A demonstrates how such a thresholded test gives rise to a curved decision boundary: to reach significance, the estimated LFC has to exceed the specified threshold by an amount that depends on the available information. We note that related approaches to generate gene lists that satisfy both statistical and biological significance criteria have been previously discussed for microarray data [23] and recently for sequencing data [19].
