Figure 9
From: Biomarker discovery in heterogeneous tissue samples -taking the in-silico deconfounding approach
Applying the deconfounding approach for classification: defining biosignatures in a simulated scenario. Classification error rates in a simulated scenario using realistic empirical noise and differentially expressed genes, which are reversely regulated in a mixture of two cell-types for a t-test-LDA approach (blue boxes) and a deconfounded-biosignature approach (red boxes). Boxplots comprise median, 1st and 3rd quartiles, as well as the 95% confidence interval (assuming normality).