Article Figures & Data
Figures
- Figure 1.
Schematic of the study design for: (A) the development and validation of the 18-hub-gene prognosis signature; and (B) the development and validation of the 12-gene predictive signature.
- Figure 2.
A, the topology of the constructed survival-related gene network in NSCLC. The gene expression of 797 survival-related genes (FDR < 10%) from 442 ADC samples in the Consortium dataset was used to construct the gene network based on the SPACE algorithm. Each node represents 1 gene (only nodes with at least 1 connection are shown). The genes with at least 7 connections with other genes were identified as hub genes and labeled in red. B, composition of the 18 survival-related hub genes. Network Connectivity refers to the number of genes that the hub gene has direct connection with based on the constructed gene network. The HR and P-values for each gene were derived from Cox models adjusted for age, cancer stage, and sample-processing sites. P-values for synthetic lethal were from our previous study (22), and P-values less than 0.05 were highlighted in yellow. Genetic alteration information was from the Tumorscape program (http://www.broadinstitute.org/tumorscape): “+” indicates the genes with significant amplification and “−” indicates significant deletion in lung cancer. The gene symbols of the 12 genes with either synthetic lethal or genetic alteration were highlighted in red.
- Figure 3.
Validation of the 18-hub-gene signature in 6 independent datasets. A, lung ADC patients; B, stage I lung ADC patients; and C, lung SCC patients. The high- and low-risk groups were defined based on the 18-hub-gene signature, which was derived from the Consortium data. The median of the estimated risk scores was used as the cut-off to partition the patients into high- and low-risk groups. Dashed gray and solid black lines indicate predicted high- and low-risk groups. Gray- and black-filled circles represent censored samples. HR compares the overall survival of the high- and low-risk group. P-values were obtained by the log-rank test. The patients with chemotherapy were excluded from the validation sets.
- Figure 4.
Comparison of the 18-hub-gene set, 18-top-ranked-gene set, and 797-SR-gene signature. A, summary of the prognostic performance for 18 hub-gene set, 18 top-ranked-gene set, and 797-SR-gene set. The training data are the Consortium data; the validation sets include 5 different datasets. aOverall HR and P-values were calculated from meta-analysis. B, expression variation across the population in the Consortium dataset based on principal component analysis. C, pair-wise mutual information distance based on expression values in the Consortium dataset. D, entropy of expression values in the Consortium dataset.
- Figure 5.
Validation of the 12-gene predictive signature in 2 independent data sets. A, JBR.10 clinical trial dataset. The high- and low-risk groups were defined by the 12-gene signature. The patients were divided into 2 equal-sized risk groups based on their estimated risk scores. In the high-risk group, patients with ACT (pink line) have significantly longer survival time than patients without ACT (observation group, blue line). In the low-risk group, patients with ACT (pink line) do not have significantly longer survival time than patients without ACT (observation group, blue line). B, UT lung SPORE dataset. The risk groups were defined by the same 12-gene signature. In the high-risk group, patients with ACT (pink line) have significantly longer survival time than patients without ACT (observation group, blue line). In the low-risk group, patients with ACT (pink line) do not have significantly longer survival time than patients without ACT.
Tables
- Table 1.
Clinical characteristics of patients in the validation datasets
SPORE New data GSE13213, Tomida (2009) GSE11969, Matsuyama (2011) GSE8894, Lee (2008) GSE3141, Bild (2006) GSE4573, Raponi (2006) GSE14814, Zhu (2010) Total patients n = 176 n = 117 n = 149 n = 138 n = 111 n = 129 n = 90 Gender Female 83 (47.2%) 57 (48.7%) 48 (32.2%) 34 (24.6%) – 47 (36.4%) 23 (25.6%) Male 93 (52.8%) 60 (51.3%) 101 (67.8%) 104 (75.4%) – 82 (63.6%) 67 (74.4%) Stage I 112 (63.6%) 79 (67.5%) 78 (52.3%) – 62 (55.9%) 73 (56.6%) 45 (50.0%) II 32 (18.2%) 13 (11.1%) 26 (17.4%) – – 33 (25.6%) 45 (50.0%) III 30 (17.0%) 25 (21.4%) 45 (30.2%) – – 23 (17.8%) – IV 1 (0.6%) – – – – – – Unknown 1 (0.6%) – – 138 (100%) 49 (44.1%) – – Histology ADCs 133 (75.6%) 117 (100%) 90 (60.4%) 62 (44.9%) 58 (52.3%) – 28 (31.1%) SCCs 43 (24.4%) – 35 (23.5%) 76 (55.1%) 53 (47.7%) 129 (100%) 52 (57.8%) Others – – 24 (16.1%) – – – 10 (11.1%) Median follow-up (months) 47.4 68 78 41.8 31.1 34.2 64.8 Platform Illumina Human-WG6 V3 Agilent 44K Agilent 21.6K custom array Affy U133 Plus_2 Affy. U133 Plus_2 Affy. U133A Affy. U133A
Supplementary Data
Files in this Data Supplement:
- Supplementary Methods - PDF file - 33K
- Supplementary Figure Legend - PDF file - 21K
- Supplementary Figure 1 - PDF file - 521K, Different expression profiles of hub genes between the lung adenocarcinoma (ADC) group and lung squamous cell carcinoma (SCC) group.
- Supplementary Figure 2 - PDF file - 200K, Summary of the prognostic performance for the 12 hub-gene set.
- Supplementary Table 1 - PDF file - 49K, Multivariate Cox regression analysis of the 18-hub-gene prognostic signature and clinical variables in the UT Lung SPORE.
- Supplementary Table 2 - PDF file - 64K, Meta-analysis of 18-hub-gene prognostic signature on ADC patients and SCC patients.
- Supplementary Table 3 - PDF file - 108K, Comparison with the gene signatures with other signatures.
- Supplementary Table 4 - PDF file - 104K, Members of the 18-hub gene signature have been included in MSigDB database C2 signature catalogues.
- SWEAVE Data - PDF file - 124K, The SWEAVE report contains all the details about the model, parameters and procedures used to generate the results in the paper.
Volume 19, Issue 6
