Skip to main content
Fig. 3 | Genome Medicine

Fig. 3

From: Transferring genomics to the clinic: distinguishing Burkitt and diffuse large B cell lymphomas

Fig. 3

Performance of the classifier trained with different BL definitions with a heatmap of Z-score normalized 28 classifier gene expression values. Classification results of GSE4732_p2, GSE10172, GSE17189 and GSE26673 when the classifier was trained by a variety of thresholds, with a heatmap of the 28 classifier genes showing the Z-score normalized expression values. The training set threshold is adjusted according to data set GSE4475 and the class probability given to each sample by the original classifier; for example, training set Th = 0.9 means only include the samples with a confidence over 0.9 in GSE4475 to train the classifier, and Strict and Wide refer to the strict and wide definition used previously. In test set GSE10172, the GEO-Class bar shows both the class label and BL probability from the original data set for each sample. The figure shows that when trained with the GSE4475 strict data set, the classifier has a strict definition of BL similar to with GSE4732_p1 but not very effective in recognizing BLs in GSE4732_p2 nor endemic BL (eBL) and HIV-related BL cases (HIV-BL GEO Gene Expression Omnibus

Back to article page