Figure 2

Multiple independent analyses were performed on the Marshfield cohort samples to assess the accuracy of subtyping by ColoType targeted RNA-seq analysis. Count data were generated by whole genome RNA-seq and by targeted RNA-seq with the ColoType custom AmpliSeq library. Gene expression values were computed from both sets of count data by size factor normalization, log2 transformed. ColoType was applied to expression data by AmpliSeq, and three independent classifiers were applied to whole-genome data. Results of the four subtype systems were compared.