Figure 1.
Outline of SCoR procedure and output results on breast cancer datasets.
(a) outline of SCoR procedure (b) number of candidate prognostic genes generated by SCoR using different resampling rounds (c) percent of candidate prognostic probesets passed the internal validation using remaining samples not used in SCoR (d) Venn diagram showing overlapping genes found by SCoR in three breast cancer datasets.
Table 1.
top overlapping prognostic genes identified from different breast cancer datasets.
Figure 2.
Identification of a common poor prognostic gene expression signature in breast cancer.
Top, cluster heatmap of gene expression using SCoR generated prognostic genes from GSE1456 dataset with a blowup of genes inside the center of the poor prognosis gene cluster (yellow box), genes matching top overlap genes from Table 1 were marked in black lines on the right side of the heatmap. Bottom panels, similar to top but only show the gene symbols from the center cluster. All genes involved in cell division cycle regulation are in black, otherwise are in grey.
Figure 3.
Comparision of NKI 70-gene signature with SCoR derived 69-gene signature.
(a), cluster heatmap of gene expression using SCoR generated prognostic probes from NKI-295 dataset with blowup views of genes inside the center of poor and good prognosis gene clusters, probes matching the NKI 70-genes were marked in black lines on the right side of the heatmap. Patient survival is marked in blue (>5yr), red (dead in 2yr), or otherwise cyan. ER status is marked in red (ER-) or cyan (ER+). Molecular subtypes are marked in blue (luminal A), green (luminal B), red (basal), ERBB2 (pink), or cyan (normal-like). NKI 70-gene prediction is marked in red (poor prognosis) or cyan (good prognosis). (b), Kaplan Meier plots of patient survivals stratified by either the NKI 70-gene signature or the SCoR 69-gene signature.
Figure 4.
Identification of a common good prognostic gene expression signature in glioblastomas.
(a) cluster heatmaps of gene expression using SCoR generated prognostic genes from TCGA and REMBRANDT datasets with annotated genes and chromosomal locations on selected good prognostic gene clusters. Patient survivals were plotted in blue (>600 days) or red (<180 days) representing top and bottom quartile survivals. PTEN copy numbers are plotted in white, green, and red, representing normal copy number, copy loss, and deletion, respectively. “topSCoR” tumors, representing a group of samples having highest expression from good prognostic genes, were marked in dark blue. (b) distribution of SCoR generated prognostic genes on all chromosomes, plotted in percentage. (c) cluster heatmap using 791 chromosome 10 probesets and 188 glioblastoma patients. PTEN copy numbers were plotted in white, green, and red, representing copy number calls of 2N, 1N, and 0N, respectively. “topSCoR” tumors marked positions of same tumors found by SCoR in panel (a). (d). histogram of percentage of chromosome 10 genes found in individual SCoR runs, blue line marked the median of these percentages and red line marked the percentage of chromosome 10 genes found by SCoR.
Figure 5.
Identification of sex as a prognostic factor in one lung adenocarcinoma dataset.
(a) cluster heatmap of gene expression using SCoR generated prognostic genes from DC2008-MI dataset annotated with gene symbols and chromosomal locations. (b) and (c) Kaplan Meier plot of patient survivals stratified by patient gender in all patients (b), and in stage T1 patients, stage T2, T3, T4 patients (c).