Comparison with Chang et al

In [1] Chang et al used two different datasets to distinguish between serum response and cycling genes. Overall there where 4962 genes in the overlap of the two experiments. Of these, Chang et al identified 42 as cell cycle regulated and 163 as serum response genes while S1 contained 230 genes in the overlap and S2 had 88 genes. As expected there was a considerable overlap between the set of genes identified as cell cycle regulated and the genes in S1. There where a total of 20 genes in the overlap (p-value < 10^-11) while there where no genes in the overlap with S2.
When looking at the set of genes identified by Chang et al as serum response genes, 9 genes where contained in S2 (significant, p-value < .001) while 12 genes where in S1 (not significant, p-value > .01). Note that while the overlap is larger than with S2, since S1 contains more than twice as many genes, 12 genes is not a significant overlap. Still, we were interested in checking this discrepancy in order to determine which of the two annotations (cell cycle by CheckSum or serum response by Chang et al) was more accurate. Of the 12 genes, four are known to be cell cycle genes (Mcm7, BRCA2, H2AFZ and Stk18) while only one (Plg) is known to be involved in inflametory response. This might indicate that the annotation provided by using static expression data is more accurate than when using time series data alone.

[1] Chang HY, Sneddon JB, Alizadeh AA, Sood R, West RB, Montgomery K, Chi JT, Rijn Mv M, Botstein D, Brown PO.
Gene Expression Signature of Fibroblast Serum Response Predicts Human Cancer Progression: Similarities between Tumors and Wounds.
PLoS Biol. 2004 Feb;2(2):E7. Epub 2004 Jan 13.