Table 1.
Characteristics of patients and samples.
Figure 1.
Correlation view of 156 samples from patients with NSCLC.
Pairwise correlations between any two samples are displayed, based on 4791 informative probe sets. The colors of the cells represent Pearson's correlation coefficient values, with deeper red indicating higher positive and deeper blue lower negative correlations. The red diagonal line displays the self-to-self comparison of each sample. Histological classification of the samples is depicted along the diagonal; the key to the color code is shown at the bottom. Histo-path_1 & Histo-path_2: initial and second histo-pathological review.
Figure 2.
Hierarchical clustering distinguishes tumors from healthy lung tissue.
A: Two-dimensional hierarchical clustering of 80 training samples, including tumors and healthy lung samples, was performed with 187 probe sets. The relative expression to the overall mean for each probe set (rows) in each sample (columns) is indicated by a color code. B: Hierarchical clustering of 156 tissue samples with 5 probe sets yields 2 groups, tumor and normal lung.
Figure 3.
Clustering analysis of NSCLC tumors with the 518 probe set histology signature.
A: agglomerative hierarchical clustering of 23 NSCLC samples using the 518 probe set histology signature. The relative expression to the overall mean for each probe set (rows) in each sample (columns) is indicated by a color code. Correlation between the samples is depicted by the dendrogram. Histo-pathological diagnosis and predictions of histology subtype by Prediction Analysis of Microarrays, using the 518 and 75 probe set signatures, are shown by colored blocks. B: correlation dendrogram generated by agglomerative hierarchical clustering of all 91 Erasmus MC NSCLC samples using the 518 probe set signature. Histo-pathological diagnosis of the initial and second review, and prediction of histology subtype by Prediction Analysis of Microarrays using the 518 probe set signature, are shown by colored blocks.
Figure 4.
Prediction of histology subtype of Erasmus MC and Duke University NSCLC samples.
A: correlation dendrogram generated by agglomerative hierarchical clustering of all 91 Erasmus MC NSCLC samples using the 75 probe set histology signature. Histo-pathological diagnosis of the initial and second review, and prediction of histology subtype by Prediction Analysis of Microarrays using the 75- and 518 probe set histology signatures, are shown by colored blocks. B: correlation dendrogram generated by agglomerative hierarchical clustering of all 96 Duke University NSCLC samples using the 75 probe set histology signature. The reported histo-pathological diagnosis, and prediction of histology subtype by Prediction Analysis of Microarrays using the 75- and 518 probe set histology signatures, are shown by colored blocks. 75AS and 518AS: prediction without the LCC probe sets in the histology signatures, using 68 and 329 probe sets respectively (see Tables S4 and S5).
Figure 5.
A 17 probe set signature predicts patient survival time.
Kaplan-Meier curves for A: 82 Erasmus MC NSCLC patients and B: 89 Duke University NSCLC patients fitted by their risk assignments based on the 17 probe set survival signature. The high- and low-risk groups differ significantly, indicated by the p-values. Grey bars indicate patients at last follow-up, still alive.
Table 2.
Multivariable proportional hazard analysis of the risk of death.
Figure 6.
Survival prediction by published prognostic signatures.
Kaplan-Meier curves for the best performing signatures (by P-value) are shown for 82 Erasmus MC patients (left) and 89 Duke University NSCLC patients (right), fitted by their risk assignments. Grey bars indicate patients at last follow-up, still alive. P-values are between brackets if overall survival of the low risk group is actually lower than that of the high risk group.
Figure 7.
Correlation view of Erasmus MC and Duke University NSCLC samples.
In total 187 tumor samples from the Erasmus MC (n = 91) and Duke University (n = 96) cohorts are shown. Pairwise correlations between any two samples are displayed, based on 3495 informative probe sets. Histological classification of the samples, and the collection source, are depicted along the diagonal. The key to the color code is shown at the bottom. Histo-path_E1 & Histo-path_E2: initial and second histo-pathological review of Erasmus MC samples. Histo-path_D: histo-pathological review of Duke University samples; Histo-path_P-518 and Histo-path_P-75: predictions by PAM of histological subtypes using the 518 and 75 probe set signatures, respectively (see Tables S4 and S5).