Figure 1
From: Ensemble analyses improve signatures of tumour hypoxia and reveal inter-platform differences
Experimental design. Outline of the experimental design for ensemble classification and evaluation of a biomarker. Microarray data is pre-processed in 24 different ways to calculate mRNA abundance levels (Stage 1). Risk groups are subsequently assigned for the evaluated biomarker (Stage 2). Each of the resulting classifications represents a vote for whether the patient is in the low or the high risk group. The ensemble score is a summation over these individual classifications and ranges from 0 to 24 (Stage 3). Only unanimously classified patients (ensemble scores 0 and 24) are considered robust and are evaluated with Cox proportional hazard ratio modeling and Kaplan-Meier survival curves (Stage 4).