OncoPage 4 of(page number not for citation purposes)BMC Bioinformatics 2009, 10:http
OncoPage 4 of(page number not for citation purposes)BMC Bioinformatics 2009, 10:http://www.biomedcentral.com/1471-2105/10/Table 1: Top significant GO groups identified between BCR/ABL and NEG phenotypes for the ALL dataset.NoGO ID GO:0043122 GO:0043123 GO:0043124 GO:0000084 GO:0000115 GO:0032715 GO:0032755 GO:0042226 GO:0045408 GO:0045410 GO:0032088 GO:0043392 GO:0043433 GO:0007257 GO:Term GW9662MedChemExpress GW9662 regulation of I-kappaB kinase/NF-kappaB … positive regulation of I-kappaB kinase/N… negative regulation of I-kappaB kinase/N… S phase of mitotic cell cycle S-phase-specific transcription in mitoti… negative regulation of interleukin-6 pro… positive regulation of interleukin-6 pro… interleukin-6 biosynthetic process regulation of interleukin-6 biosynthetic… positive regulation of interleukin-6 bio… inhibition of NF-kappaB transcription fa… negative regulation of DNA binding negative regulation of transcription fac… activation of JNK activity positive regulation of JNK activityRanka 3 -p-value 3.7e-2.9e-0.0.0.a Refers to the rank on the list of top GO terms enriched by individual GO term analysis for FDR controlled at 5 . The dash (-) indicates that this GO term was not found on the list generated by individual GO term analysis.gene activates Jun kinase and requires Jun for transformation [29]. The accuracy of the enriched GO cluster method for interleukin-6 was also supported by previous research [30]. The disadvantage with parent-based enrichment studies is that they use only the most recent parent node to calculate the significance of gene sets. Therefore, we grouped the GO terms into a cluster, which was a feasible solution for improving the sensitivity of our programme. For example, Figure 3 shows a subgraph induced by S phase of mitotic cell cycle: 0000084 regulation of transcription during PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/27484364 S-phase of mitotic cell cycle: 0000115 that exhibited significant differential expression by CeaGO. In this subgraph, the Benjamini and Yekutieli adjusted p-value of PAGE analysis of S phase of mitotic cell cycle: 0000084 using all genes with GO terms was only 0.025 (p-value = 0.0068 in elim), which was much less significant than the result of CeaGO analysis (p-value = 2.9e-5). Some groups returned similar results to the parent-based method, however. For example, activation of JUN kinase activity: 0007257 positive regulation of JUN kinase activity: 0043507 had an adjusted pvalue of 0.045, while the parent node positive regulation of JUN kinase activity: 0043507 had an adjusted p-value of 0.038. In these cases, cluster members identified by CeaGO were almost the same as the offspring of the GO term enriched by parent-based analysis, which lead to identical significant score. These results indicate that the CeaGO method is more sensitive at detecting certain novel expression changes than parent-based enrichment methods, while some coordinated changes were preserved.Application of CeaGO to the ALL/AML dataset The purpose of the present study was to examine the applicability of the CeaGO algorithm. In addition, we tested the algorithm with a published dataset called golubEsets [14], which is also available from Bioconductor. It consists of 7,129 genes from 47 samples of acute lymphoblastic leukemia (ALL), and 25 samples of acute myeloblastic leukemia (AML). Normalization on these samples was also carried out using VSN. This pre-processing resulted in 6,372 genes annotated to GO terms from BP ontology. The induced GO graph contains 2,766 GO t.
