Sebastian Hoersch

Gene Expression Profiles Predict Early Relapse in Ovarian Cancer after Platinum-Paclitaxel Chemotherapy

Purpose: Women with advanced epithelial ovarian cancer are routinely treated with platinum-paclitaxel chemotherapy following cytoreductive surgery, yet only ∼20% achieve long-term disease-free survival. We hypothesized that differences in gene expression before treatment could distinguish patients with short versus long time to recurrence after administration of platinum-paclitaxel combination chemotherapy. Experimental Design: To test this hypothesis, gene expression profiling of 79 primary surgically resected tumors from women with advanced-stage, high-grade epithelial ovarian cancer was done using cDNA microarrays containing 30,721 genes. Supervised learning algorithms were applied in an effort to develop a binary classifier that could discriminate women at risk for early (≤21 months) versus late (>21 months) relapse after initial chemotherapy. Results: A 14-gene predictive model was developed using a set of training samples (n = 51) and subsequently tested using an independent set of test samples (n = 28). This model correctly predicted the outcome of 24 of the 28 test samples (86% accuracy) with 95% positive predictive value for early relapse. Conclusions: Predictive markers for early recurrence can be identified for platinum-paclitaxel combination chemotherapy in primary ovarian carcinoma. The proposed 14-gene model requires further validation.

Comparison of the Predictive Accuracy of DNA Array-Based Multigene Classifiers across cDNA Arrays and Affymetrix GeneChips

We examined how well differentially expressed genes and multigene outcome classifiers retain their class-discriminating values when tested on data generated by different transcriptional profiling platforms. RNA from 33 stage I-III breast cancers was hybridized to both Affymetrix GeneChip and Millennium Pharmaceuticals cDNA arrays. Only 30% of all corresponding gene expression measurements on the two platforms had Pearson correlation coefficient r >or= 0.7 when UniGene was used to match probes. There was substantial variation in correlation between different Affymetrix probe sets matched to the same cDNA probe. When cDNA and Affymetrix probes were matched by basic local alignment tool (BLAST) sequence identity, the correlation increased substantially. We identified 182 genes in the Affymetrix and 45 in the cDNA data (including 17 common genes) that accurately separated 91% of cases in supervised hierarchical clustering in each data set. Cross-platform testing of these informative genes resulted in lower clustering accuracy of 45 and 79%, respectively. Several sets of accurate five-gene classifiers were developed on each platform using linear discriminant analysis. The best 100 classifiers showed average misclassification error rate of 2% on the original data that rose to 19.5% when tested on data from the other platform. Random five-gene classifiers showed misclassification error rate of 33%. We conclude that multigene predictors optimized for one platform lose accuracy when applied to data from another platform due to missing genes and sequence differences in probes that result in differing measurements for the same gene.