Table 2. Statistical comparison of gene expression similarity (gene similarity scores).
Dataset | Compared genes to background | Mann-Whitney U-Test | Permutation Test | Top Scored Imprinted Genes | Hyper-geometric Test |
---|---|---|---|---|---|
GSE6506 [19] | Pluripotent | 0.006 | 0.019 | 55 | 0.006 |
Hematopoietic | 0.044 | 0.032 | 57 | 0.010 | |
GSE34723 [21] | Pluripotent | 0.004 | 0.022 | 50 | 0.004 |
Hematopoietic | 0.003 | 0.006 | 51 | 0.009 | |
GSE14833 [20] | Pluripotent | 0.003 | 0.014 | 18 | 0.195 * |
Hematopoietic | 0.006 | 0.072 * | 24 | 0.214 * | |
GSE10246 [22] (Control) | Pluripotent | 0.106 | 0.267 | 11 | 0.784 |
Hematopoietic | 0.101 | 0.089 | 14 | 0.700 |
First, the Mann-Whitney U-test was used to test whether imprinted genes have a higher gene expression similarity to pluripotent and hematopoietic genes than to the background of all other genes (non-imprinted genes). Then the top 10% scoring genes were tested using hyper-geometric test to find out the significance of having imprinted genes among them (rightmost column). Secondly, also a permutation test was performed to test the null distribution hypothesis by randomly shuffling the expression values of the imprinted genes and recalculating the similarity scores. The procedure was repeated 1000 times. The p-value was computed based on the number of random times where the similarity scores were higher than the real score without shuffling. (*) Among the three hematopoietic datasets, only the p-value of the hyper-geometric test for GSE14833 and the respective permutation test does not meet the significance threshold of 0.05.