Figure 1

Description of geNetClassifier main features and results. (A) Scheme representing a gene-disease space for three hypothetical diseases. The color-line ovals would enclose the genes (dots) affected by a given disease. The coloured-background circles would mark the altered genes that are specific for a disease, which are the ones we aim to identify: i.e. genes affected by only one of the compared diseases. (B) Graph presenting the posterior probability of the top 2000 genes included in the gene ranking for each of the 4 leukemia subtypes and the non-leukemia samples. The genes are selected and ranked based on their posterior probability for each class. Genes with posterior probability over the threshold (> 0.95) can be considered significant candidates to mark each disease subtype. (C) Lists presenting the top-15 genes in the ranking of each class. The bottom row shows the total number of significant genes in the whole ranking. The shaded area contains the genes selected as the minimum subset to separate the classes. The number of genes selected per class is shown on top. (D) Discriminant power plot of the first ranked gene for each leukemia subtype: VPREB1 (pre-B lymphocyte 1) for ALL; HOXA9 (homeobox 9) for AML; TYMS (thymidylate synthetase) for CLL; and GJB6 (gap junction beta-6 30 kDa protein) for CML. The red numbers indicate the discriminat power values assigned to each gene.