Results for the yeast dataset

We have downloaded raw expression data from the Stanford Microarray Database. We have downloaded the values for the two channels (synchronized and unsynchronized samples) and computed their ratio for each time point. Using a p-value of 0.01 Checksum identified 52 genes in this dataset. These were divided to two sets: The first, S1, contained the 21 genes that were expressed at lower than expected levels. The second S2 contained the 31 genes expressed at higher than expected levels. 12 of the genes in S1 where also included in the 800 genes identified by Spellman et al to be cycling in yeast (p-value = 3*10^(-7)). Below we plot the expression curves for these 12 genes. Next, we looked at S2. While 18 of the 21 orfs (86%) identified in S1 were annotated in GO, only 14 of the 31 orfs (45%) identified for S2 were annotated. Most of the S2 orfs where either "transposable_element_gene" or classified as "ORF, Dubious" in GO. This indicates that the expression seen for the array spots which correspond to these orfs are likely artifacts (or noise). Checksum was able to correctly identify such features since their expression, in the average case, was low while noise in the time series data made them look more active.
12 S1 genes determined to be cycling