Skip to main content
. Author manuscript; available in PMC: 2013 Dec 19.
Published in final edited form as: N Am J Med Sci (Boston). 2013 Jul 25;6(3):10.7156/najms.2013.0603107. doi: 10.7156/najms.2013.0603107

Fig. 2. Performance of “classifier” genes for ASD subtypes vs. control samples.

Fig. 2

The differentially expressed genes used for class prediction were selected by USC with 10-fold cross-validation. SVM analyses of microarray expression data for the selected genes show: A) Separation of severely language-impaired ASD (L, red) from controls (C, turquoise) based on 29 genes; B) Separation of mild ASD (M, blue) from controls (C ) based on 27 genes; C) Separation of ASD with savant skills (S, yellow) from controls (C ) based on 18 genes; D) Separation of combined ASD samples (each subtype represented by its respective color) from controls.