Fig. 1.

Flow-chart of the procedure followed in the pre-processing and analysis of the data. Six datasets (1 FF and 5 FFPE, each with 54 samples and 18 genes) underwent quality control procedures before analysis. Thirty-nine [39] “good” samples and 16 “good” genes were retained. Correlation analyses were performed using mean scores from the sample pairs. The predictive ability of the 16–gene set was validated using the Affymetrix FF, Affymetrix FFPE and NanoString gene expression data, by the PAM method