Fig. 1

Comparison of normalized ranks predicted by Scuba and competing kernel methods. Normalized test genes rank distributions predicted by Scuba, MKL1class and ProDiGe for test genes in (a) genome-wide prioritizations in the unbiased evaluation of Table 2 - (b) candidate set-based prioritizations in the unbiased evaluation of Table 2 - (c) genome-wide prioritizations in the expanded unbiased evaluation of Table 4. In all cases, each point represents a test gene and lower values on the axes indicate better predictions. Genes lying on a diagonal have the same rank according to both methods considered on a plot. The further a gene lies above (below) a diagonal and the better it was ranked by Scuba (MKL1class/ProDiGe) compared to MKL1class/ProDiGe (Scuba). In each plot we show the Pearson correlation coefficient r between the test genes rank distributions and its associated p-value