Co-expression Extrapolation Research Articles

A gene expression algorithm for the prediction of platinum sensitivity in ovarian cancer.

5021 Background: Despite advances in radical surgery and chemotherapy delivery, ovarian cancer is the most lethal gynecologic malignancy. In vitro models predicting chemosensitivity rely on cellular phenotypes and are hampered by the extrapolation of data from culture to clinic. Gene expression models can predict therapeutic responses without these constraints. We tested our gene based co-expression extrapolation (COXEN) algorithm on a cohort of 65 ovarian cancer patients to assess its clinical utility. Methods: COXEN is an in silico algorithm that predicts chemosensitivity of tumors based on genomic expression signatures and has been reported previously. In brief, NCI- 60 cell lines were used to identify candidate drug sensitivity biomarkers for carboplatin and paclitaxel. Candidate biomarkers were then narrowed by examining concordant expression patterns between cell lines and a historical set of ovarian cancer patients. Biomarkers were further triaged by examining expression patterns of frozen and formalin-fixed paraffin embedded (FFPE) tissue samples. Models were refined using an independent cohort from which the final prediction models were obtained. To evaluate the clinical utility of COXEN in ovarian cancer, we performed genome-wide profiling on FFPE samples of 65 patients obtained prior to adjuvant chemotherapy. We then retrospectively applied COXEN to this cohort to evaluate the predictive success. Results: COXEN correctly stratified platinum sensitive and resistant patients (p = 0.019) with sensitivity = 93%, specificity = 33%, PPV = 65%, and NPV = 78%. Paclitaxel response was also significantly stratified (p = 0.033) with sensitivity = 96%, specificity = 26%, PPV = 61%, and NPV = 86%. Models for platinum-taxane combination demonstrated a significant survival difference between the predicted responders and nonresponders with median survival of 12.9 months vs. 8.1 months (p = 0.045). Conclusions: COXEN successfully predicted platinum resistance and reliably discriminated taxane response in this retrospective cohort. Prediction of platinum response is clinically relevant and, if reliable, could alter primary management of ovarian cancer. Prospective evaluation of the COXEN model is underway. Author Disclosure Employment or Leadership Position Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Key Genomics Key Genomics

Abstract 2630: Use of the “coexpression extrapolation” algorithm to identify tumor biomarkers that predict drug sensitivity to Vorinostat and Velcade in non-small cell lung cancer

Abstract Background: Non-small cell lung cancer (NSCLC) is the leading cause of death from cancer in the United States. Research is currently focused on identifying novel molecularly-targeted therapies that can improve on the rather dismal results of platinum-based doublet therapies. Additionally, the inability to accurately predict the efficacy of these agents for each individual patient remains a problem with the use of molecularly-targeted therapies. The purpose of this study was to identify tumor biomarkers using the “coexpression extrapolation (COXEN)” algorithm that predict drug sensitivity to Vorinostat, a histone deacetylase inhibitor, and Velcade, a proteasome inhibitor, two molecularly-targeted agents currently in a Phase I clinical trial for NSCLC. Methods: Applying COXEN, biomarker prediction models were first discovered and trained for Vorinostat and Velcade based on in vitro drug activity profiles of the 9 NSCLC cell lines (NCI-9) in the National Cancer Institute's public database of 60 cancer cell lines. Independently, a panel of 40 NSCLC cell lines (UVA-40) was experimented with either Vorinostat or Velcade treatments in a dose-escalation schema to obtain 50% growth inhibition (GI50) values. Genome-wide expression profiles for both the NCI-9 and UVA-40 cell lines were available or obtained with Affymetrix HG-U133A GeneChips®. Prospective predictability of the two drugs' COXEN models was confirmed with 100 random splits of UVA-40 for robust biomarker selection and independent evaluation of COXEN prediction models. The final COXEN models of the two drugs were determined using all UVA-40 lines. Results: The COXEN biomarker models with 45 genes for Vorinostat and 15 genes for Velcade prospectively provided drug sensitivity prediction scores that were highly significantly associated with the drug activity GI-50 values of Vorinostat (p=0.006) and Velcade (p=0.002) on the UVA-40, respectively. Analysis of the robustness of the biomarker selection for each drug was also highly significant (p<0.005), underscoring a high fidelity of our biomarker selection and prediction modeling. Finally, review of selected biomarkers for both drugs revealed no significant overlap with only one shared biomarker (CFLAR) for both Vorinostat and Velcade, implying quite independent modes of mechanisms of the two drugs. Conclusions: Biomarker models generated through the in vitro-based COXEN algorithm were highly predictive of tumor sensitivity to Velcade and Vorinostat in NSCLC cell lines. COXEN genomic biomarker models may significantly help predict a priori the efficacy of novel targeted therapeutics Velcade and Vorinostat for NSCLC patients. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 2630.

Abstract 2700: In vitro-based COXEN biomarker model predicts outcome in metastatic colorectal cancer patients treated with cetuximab

Abstract Purpose While metastatic colorectal cancer (CRC) patients with KRAS mutations have shown extremely low clinical responses to cetuximab, an epidermal growth factor receptor (EGFR) inhibitor, about half of metastatic CRC patients with KRAS wild type have been found to be responsive to this targeted therapeutic agent. The current study was performed to apply refined “coexpression extrapolation” (COXEN) algorithm in order to identify and develop a predictive gene expression signature of cetuximab for metastatic CRC patients using in vitro drug activity data of CRC cell lines treated with erlotinib, another EGFR inhibitor. Patients Gene expression profiles (Illumina HumanHT-12 BeadChips®) and KRAS mutation information were obtained for 42 metastatic CRC patients at the Department of Medical School, University Paris Descartes. All patients received either a cetuximab monotherapy or a cetuximab-based combination regimen as a second-line or greater therapy. KRAS status of each patient's tumor was assessed by allelic discrimination. Each patient's clinical response to therapy was measured according to RECIST (Response Evaluation Criteria in Solid Tumors) criteria. Methods The refined COXEN algorithm comprising three main steps was applied to construct a multivariate prediction model of cetuximab for metastatic CRC patients. First, the gene expression profiles (Affymetrix HG-U133Plus2 GeneChips®) and the in vitro drug activities on erlotinib, which were previously published on 8 CRC cell lines, were utilized to identify drug sensitivity biomarkers most highly associated with the drug activities. Using these biomarkers, candidate multivariate prediction models were developed using the 8 CRC cell lines and evaluated on 7 CRC cell lines from the NCI-60 cell lines treated with erlotinib. The final prediction signature model with 70 biomarkers was prospectively validated on a CRC patient set recently reported in Khambata-Ford et al. (2007). For our patient set, a smaller 60-gene biomarker model was established by annotation matches between the two different array platforms. Results The refined COXEN signature models accurately predicted the clinical responses of metastatic CRC patients with KRAS wild-type receiving the cutuximab single or combination therapy (Spearman correlation test p-values between predicted scores and different degrees of clinical responses = 0.005 and 0.034 for Khambata-Ford and our patient sets, respectively). We also found that the COXEN prediction scores for patients were highly correlated with their disease-free survival times (Spearman correlation test p-values = 0.01 and 0.12 for the two sets, respectively). Conclusion The refined COXEN biomarker signature models demonstrated high potential and clinical utility to select a priori metastatic CRC patients who are likely to benefit from treatments with cetuximab. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 2700.

The COXEN Principle: Translating Signatures of In vitro Chemosensitivity into Tools for Clinical Outcome Prediction and Drug Discovery in Cancer

Substantial effort has been devoted to in vitro testing of candidate chemotherapeutic agents. In particular, the United States National Cancer Institute Developmental Therapeutics Program (NCI-DTP) Human Tumor Cell Line Screen has screened hundreds of thousands of compounds and extracts, for which data on more than 40,000 compounds tested on a panel of 60 cancer cell lines (NCI-60) are publically available. In tandem, gene expression profiling has brought about a sea change in our understanding of cancer biology, allowing discovery of biomarkers or signatures able to characterize, classify, and prognosticate clinical behavior of human tumors. Recent studies have used tumor profiling matched to clinical trial outcome data to derive gene expression models predicting therapeutic outcomes, though such efforts are costly, time-consuming, tumor type-specific, and not amenable to rare diseases. Furthermore, addition of new or established drugs to multidrug combinations in which such models are already available requires the entire model to be re-derived. Can the aforementioned in vitro testing platform, coupled to the universal language of genomics, be used to develop, a priori, gene expression models predictive of clinical outcomes? Recent advances, including the CO-eXpression ExtrapolatioN (COXEN) algorithm, suggest that development of these models may be possible and raise important implications for future trial design and drug discovery.

Prospective Comparison of Clinical and Genomic Multivariate Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer

Several different multivariate prediction models using routine clinical variables or multigene signatures have been proposed to predict pathologic complete response to combination chemotherapy in breast cancer. Our goal was to compare the performance of four conceptually different predictors in an independent cohort of patients. Gene expression profiling was done on fine-needle aspirations of 100 stage I to III breast cancers before preoperative paclitaxel, 5-fluorouracil, doxorubicin, and cyclophosphamide combination chemotherapy. Pathologic response was correlated with prediction results from a clinical nomogram, a human cancer-derived genomic predictor (DLDA30), a cell line-based genomic predictor [in vitro coexpression extrapolation (COXEN)], and an optimized cell line-derived (in vivo COXEN) predictor. None of the 100 test cases were used in the development of these predictors. The in vitro COXEN using a combination of four individual drug sensitivity predictions derived from cell lines was not predictive [area under the receiver operator characteristic curve (AUC), 0.5; 95% confidence interval, (95% CI), 0.41-0.59]. The clinical nomogram (AUC, 0.73; 95% CI, 0.65-0.80) and the DLDA30 (AUC, 0.73; 95% CI, 0.66-0.80) genomic predictor had similar performances. The in vivo COXEN that used informative genes from cell lines but was trained on a separate human data set also showed significant predictive value (AUC, 0.67; 95% CI, 0.60-0.74). These three different prediction scores correlated with each other and were significant in univariate but not in multivariate analysis. Three conceptually different predictors performed similarly in this validation study and tended to identify the same patients as responders. A genomic predictor that relied solely on a composite of individual drug sensitivity predictions from cell lines did not show any predictive value.

A strategy for predicting the chemosensitivity of human cancers and its application to drug discovery

The U.S. National Cancer Institute has used a panel of 60 diverse human cancer cell lines (the NCI-60) to screen >100,000 chemical compounds for anticancer activity. However, not all important cancer types are included in the panel, nor are drug responses of the panel predictive of clinical efficacy in patients. We asked, therefore, whether it would be possible to extrapolate from that rich database (or analogous ones from other drug screens) to predict activity in cell types not included or, for that matter, clinical responses in patients with tumors. We address that challenge by developing and applying an algorithm we term "coexpression extrapolation" (COXEN). COXEN uses expression microarray data as a Rosetta Stone for translating from drug activities in the NCI-60 to drug activities in any other cell panel or set of clinical tumors. Here, we show that COXEN can accurately predict drug sensitivity of bladder cancer cell lines and clinical responses of breast cancer patients treated with commonly used chemotherapeutic drugs. Furthermore, we used COXEN for in silico screening of 45,545 compounds and identify an agent with activity against human bladder cancer.