Prediction analysis of microarrays (PAM) applied to the data set of clinical samples and representative cell lines. Misclassification errors were determined by using 10-fold cross-validation, as described previously.¹²  (A) Misclassification error of the expression signatures (shrunken centroids) as a function of the number of genes used in the classifier. (B) Misclassification error of the expression signatures (shrunken centroids) for each of 3 subtypes of ALL as a function of the number of genes used in the classifier. (C) Expression signatures (shrunken centroids) for the 3 different subtypes using a threshold = 3.0 (270 clones representing 235 genes).