Table 2. Most significantly enriched GO terms and KEGG pathways.

A) Tests of 507 Geneontology (GO) biological process terms with at least 10 genes present on the arrays.
Comparison	direction	number of probe sets selected (of 22690)	number of distinct genes selected (of 13142)	GO term title	number of genes with GO term on the array	number of those genes selected	p-value
Tmsc vs Tn	up in Tmsc	672	543	cell cycle	335	41	4.1E-10
				immune response	263	29	1.4E-06
				DNA replication	96	16	1.8E-06
				mitosis	109	17	2.2E-06
				cell division	173	22	2.6E-06
				regulation of progression through cell cycle	185	20	8.1E-05
Tmsc vs Tn	up in Tn	545	448	transcription	1125	72	1.1E-07
				regulation of transcription, DNA-dependent	1486	86	4.6E-07
Te vs Tmsc	up in Te	1486	1098	cell cycle	335	102	1.5E-32
				mitosis	109	53	1.1E-28
				cell division	173	66	1.1E-27
				DNA replication	96	40	6.7E-19
				apoptosis	296	54	3.1E-08
				DNA replication initiation	13	9	1.0E-07
Te vs Tmsc	up in Tmsc	1372	1036	ribosome biogenesis and assembly	47	21	9.6E-12
				translation	259	51	7.5E-10
Te vs Tn	up in Te	1817	1359	cell cycle	335	109	2.6E-29
				mitosis	109	52	2.9E-23
				cell division	173	67	3.5E-23
				DNA replication	96	41	1.8E-16
				DNA replication initiation	13	10	2.9E-08
				apoptosis	296	59	5.1E-07
				DNA recombination	35	13	2.4E-05
				regulation of progression through cell cycle	185	38	2.6E-05
Te vs Tn	up in Tn	1837	1359	ribosome biogenesis and assembly	47	18	5.3E-07
				translation	259	49	3.3E-05
				transcription	1125	158	6.9E-05

B) Tests of 190 KEGG pathways
Comparison	direction	number of probe sets selected (of 22690)	number of distinct genes selected (of 13142)	Pathway Title	number of genes in pathway on the array	number of those genes selected	p-value
Tmsc vs Tn	up in Tmsc	672	543	Cell cycle	102	16	4.1E-06
				Pyrimidine metabolism	80	12	1.0E-04
				Jak-STAT signaling pathway	122	14	5.0E-04
Tmsc vs Tn	up in Tn	545	448	Wnt signaling pathway	133	11	5.8E-03

We fit a 1-way ANOVA model to the data for each probe set and used it to compare each pair of groups. We selected those probe sets that gave p-values of p<.01 and average fold differences of at least 1.5-fold as being differentially expressed. We estimated the false discovery rate (FDR) by performing an identical analysis of all data sets with permuted sample labels that did not recapitulate the actual groupings of the 3 groups of samples, and dividing the average number of probe sets selected in these permuted data sets by the actual number of probe sets selected for the original data, and obtained estimated FDRs of 7.5% for T_MSC vs. T_N, 3.1% for T_E vs. T_MSC, and 2.5% for T_E vs. T_N. The number of probe sets selected as being increased or decreased is shown, as well as the number of distinct genes these represented, determined using Entrez Gene identifiers. The mapping of probe sets to genes, as well as the Gene Ontology terms for each gene, were obtained from Affymetrix web site, and were dated July 11, 2007. We tested each list of distinct genes for over-representation in 507 lists of Gene Ontology terms for which there were at least 10 genes present on the array, as well as for 190 lists of pathways from the Kyoto Encyclopedia of Genes and Genomes, obtained Jun 12, 2007 (http://www.genome.jp/kegg/). We present only the most significant lists obtained in each category, and of the most significant lists present all of them (that is, none are hidden). The total number of genes in each list that were on the array is given as well as the number of those that were selected as altered in each pair-wise comparison of groups of samples. The p-values shown are from using these counts to perform one-sided Fisher's Exact tests, however, since many lists were tested, we also give the average number of lists that obtained a p-value this smaller or smaller in 100 analyses of the same data in which the gene labels were randomly permuted.