Therefore, we suggest that the predictive performance of the 14-gene ERBB2 signature is Bortezomib manufacturer better than that of the single ����best probe set����. We tested the predictive performance of the 14-gene signature in 2 validation sets . The first validation set is composed of 278 breast tumor profiles; the prediction accuracy was 94.60%, sensitivity 76.27%, specificity 99.54%, PPV 97.83% and NPV 93.97% . For the second validation set , the prediction accuracy was 93.55%, sensitivity 83.07%, specificity 98.39%, PPV 96.30% and NPV 92.42% . Importantly, the second validation set was obtained from transcript profiles performed on a different type of GeneChip �C HG-U133 Plus 2.0. We performed this last validation on data collected from HG-U133 Plus 2.0 GeneChips to determine whether the candidate 14-gene ERBB2 signature was capable of separating ERBB2-positive tumors from their ERBB2- negative counterparts WZ4002 citations independent of the nature of the Affymetrix arrays to which the transcripts were hybridized. Figure 4 and Table S4 depict sensitivity and specificity levels obtained for the training and the validation sets using the 14-gene ERBB2 signature or using the method employing a single probe set . The specificity levels obtained by using one probe set were relatively high, ranging between 94.94% and 99.54% ; however, the sensitivity levels were significantly lower, ranging between 54.55% and 77.78% . Whereas the specificity levels were approximately within the same range using the 14- gene ERBB2 signature, the sensitivity levels changed to range between 59.09% and 77.78% . Impor-tantly, the sensitivity and specificity obtained with HG-133 Plus 2 array lie within the 95% confidence interval for both sensitivity and specificity obtained for HG-U133A arrays, for which our 14-gene ERBB2 signature was originally developed. Global gene expression profiling is widely used in cancer research and the results of these analyses are generally accessible to the scientific community in public repositories. However, these profiles rarely have accessory information concerning the clinically established status of PR, ER or that of ERBB2. Knowledge of the expression of the aforementioned markers could be used to mine publically available gene expression profiles for candidate molecular targets thus aiding efforts to expand the armamentarium of anticancer therapies targeted to these breast tumor subtypes. Previous studies have demonstrated a correlation between mRNA levels and clinical receptor status as established by IHC, FISH and ligand-binding assays using breast tumor samples . Means have also been established for statistical thresholds for ESR1, PR and ERBB2 transcript levels to assign their expression status in profiled breast tumor samples .