Figure 1.

Linear discriminant analysis applied to microarray data for 318 genes discriminant between normal and tumoural prostate samples generates minimal-size gene signatures diagnostic of PCa. Results of the best models generated by each of the LDA approaches used: (A) backward stepwise deterministic, (B) backward stepwise stochastic and (C) forward stepwise. Plots show the performance of the models for the 65 tumoural (filled circles) and 19 normal (empty circles) samples analysed. LDA scores obtained by each sample are represented on the y axis. Samples are classified as tumoural or non-tumoural, when their corresponding LDA score are above or under the model threshold (dashed line), respectively. Note that each LDA model yields its own independent class-discriminant threshold.