TY - JOUR
T1 - Sharpening Pathway-Based Analyses using Hierarchical Clustering: Results of a Prostate Cancer Genome-Wide Association Study
AU - Menashe, Idan
AU - Maeder, Dennis
AU - Chanock, Stephen
AU - Rosenberg, Philip S.
AU - Chatterjee, Nilanjan
PY - 2010/12
Y1 - 2010/12
N2 - Pathway analyses compliment primary scans of genome-wide association studies (GWAS) to indentify overrepre-sentation of susceptibility loci in predefined gene-sets.We applied two pathway analysis methods: gene-setenrichment analysis (GSEA), and adaptive rank-truncatedproduct (ARTP) to the Cancer Genetic Markers of Suscept-ibility (CGEMS) prostate cancer GWAS (1172 cases and 1157control), using pathways with 10-100 genes from threepublically available resources (BioCarta, KEGG, and PID).An artificial pathway composed of 31 genes associated withprostate cancer was used as a positive control. Pathwayswith FDRo0.2 were considered noteworthy.A high concordance was seen between results of GSEA andARTP (R 5 0.72; P0). Of the 445 pathways tested, only the‘ ‘Action of Nitric Oxide in the Heart’’ pathway from BioCartawas noteworthy using ARTP (p 5 0.0002; FDR 5 0.09). Thepositive control pathway was ranked 2nd using AR TP(p 5 0.0027; FDR 5 0.60), but only 12th using GSEA(p 5 0.0182; FDR 5 0.54). Next, we used hierarchical cluster-ing to merge pathways with similar gene content and reducethe number of gene-sets to 167. Post clustering, one gene-setswasnoteworthybybothGSEAandARTP(p 5 0.0008;FDR 5 0.13, and p 5 0.001; FDR 5 0.18 respectively), andthree other gene-sets wer e noteworthy by ARTP only.These results suggest that gene-set clustering may improvepower of both GSEA and ARTP, and facilitate detection ofpathways underlying prostate cancer predisposition.
AB - Pathway analyses compliment primary scans of genome-wide association studies (GWAS) to indentify overrepre-sentation of susceptibility loci in predefined gene-sets.We applied two pathway analysis methods: gene-setenrichment analysis (GSEA), and adaptive rank-truncatedproduct (ARTP) to the Cancer Genetic Markers of Suscept-ibility (CGEMS) prostate cancer GWAS (1172 cases and 1157control), using pathways with 10-100 genes from threepublically available resources (BioCarta, KEGG, and PID).An artificial pathway composed of 31 genes associated withprostate cancer was used as a positive control. Pathwayswith FDRo0.2 were considered noteworthy.A high concordance was seen between results of GSEA andARTP (R 5 0.72; P0). Of the 445 pathways tested, only the‘ ‘Action of Nitric Oxide in the Heart’’ pathway from BioCartawas noteworthy using ARTP (p 5 0.0002; FDR 5 0.09). Thepositive control pathway was ranked 2nd using AR TP(p 5 0.0027; FDR 5 0.60), but only 12th using GSEA(p 5 0.0182; FDR 5 0.54). Next, we used hierarchical cluster-ing to merge pathways with similar gene content and reducethe number of gene-sets to 167. Post clustering, one gene-setswasnoteworthybybothGSEAandARTP(p 5 0.0008;FDR 5 0.13, and p 5 0.001; FDR 5 0.18 respectively), andthree other gene-sets wer e noteworthy by ARTP only.These results suggest that gene-set clustering may improvepower of both GSEA and ARTP, and facilitate detection ofpathways underlying prostate cancer predisposition.
U2 - 10.1002/gepi.20553
DO - 10.1002/gepi.20553
M3 - תקציר הצגה בכנס
SN - 0741-0395
VL - 34
SP - 939
EP - 939
JO - Genetic Epidemiology
JF - Genetic Epidemiology
IS - 8
ER -