Figure-3:

Bar chart and table showing 4-class accuracy of CMS classifiers, along with number (proportion) of samples classified to each CMS. We assess accuracy for classifier with top 100 genes in terms of FF/FFPE correlation for FFPE and FF, computed based on Nanostring measurements for FFPE and FF in current study (Nano FFPE-100, Nano FF-100) and based on Affymetrix measurements for FF in the Affy CRCSC validation data set (V2a, Affy FF-100), and for the full 472 gene classifier applied to FF samples run on Nanostring platform for current study (Nano FF-472) and FF samples run on Affymetrix in the Affymetrix CRCSC validation data set (Affy FF-472). Performance is summarized overall and for subsets of samples with high classification confidence (α>0.50, 0.80 or 0.90).