Skip to main content
Figure 6 | BMC Medical Genomics

Figure 6

From: Using logistic regression to improve the prognostic value of microarray gene expression data sets: application to early-stage squamous cell carcinoma of the lung and triple negative breast carcinoma

Figure 6

Genes predicting early death versus long survival with the revolving sliding window method. The scores shown on the Y-axis represent the number of windows in which an accuracy (area under a ROC plot) greater than 0.8 was achieved by logistic regression. The genes most frequently exceeding the 0.8 threshold are shown in order of numbers of occurrences out of the maximum possible score of 80, which is the total number of comparisons. (a) As shown in Figure 5, 40 sequential windows of 20 cases each were compiled from the early death group of 40 (Figure 2) and compared to a constant set of the 20 longest term survivors; thus, the maximum possible score was 40. Only genes which achieved a score > 10 of 40 are shown. (b) Revolving sliding windows of 20 long term survival cases each among the group of 40 longest survivors were compared to a constant set of 20 early deaths (Figure 2). Again, the 40 comparisons allowed a maximum score of 40. (c) Summation of the results from panels A and B. Because a total of 80 comparisons were made, the maximum possible score was 80. The two most accurate genes (CD79A and CD27) achieved scores approaching 50.

Back to article page