A 20-gene model for molecular nodal staging of bladder cancer: development and prospective assessment.

BACKGROUND: Neoadjuvant chemotherapy before cystectomy confers a survival benefit in bladder cancer, but it has not been widely adopted since most patients do not benefit and we are at present unable to predict those that do. Since the most important predictor of recurrence after cystectomy is pathologically positive nodes, our aim was to assess techniques that define this stage for the selection of patients for neoadjuvant chemotherapy. METHODS: We developed a gene expression model (GEM) to predict the pathological node status in primary tumour tissue from three independent cohorts of patients who were clinically node negative. From a subset of transcripts detected faithfully by microarrays from both paired frozen and formalin-fixed tissues (32 pairs), we developed both the GEM and cutoffs that identified patient strata with raised risk of nodal involvement by use of two separate training cohorts (90 and 66 patients). We then assessed the GEM and cutoffs to predict node-positive disease in tissues from a phase 3 trial cohort (AUO-AB-05/95; 185 patients). FINDINGS: We developed a 20-gene GEM with an area under the curve of 0·67 (95% CI 0·60-0·75) for prediction of nodal disease at cystectomy in AUO-AB-05/95. The cutoff system identified patients with high relative risk (1·74, 95% CI 1·03-2·93) and low relative risk (0·70, 95% CI 0·51-0·96) of node-positive disease. Multivariate logistic regression showed the GEM predictor was independent of age, sex, pathological stage, and lymphovascular space invasion (coefficient 9·81, 95% CI 1·64-18·00; p=0·019). INTERPRETATION: Selecting patients for neoadjuvant chemotherapy on the basis of risk of node-positive disease has the potential to benefit high-risk patients while sparing other patients toxic effects and delay to cystectomy. FUNDING: US National Cancer Institute (R01CA143971).

Network Rewiring in Cancer: Applications to Melanoma Cell Lines and the Cancer Genome Atlas Patients.

Genes do not work in isolation, but rather as part of networks that have many feedback and redundancy mechanisms. Studying the properties of genetic networks and how individual genes contribute to overall network functions can provide insight into genetically-mediated disease processes. Most analytical techniques assume a network topology based on normal state networks. However, gene perturbations often lead to the rewiring of relevant networks and impact relationships among other genes. We apply a suite of analysis methodologies to assess the degree of transcriptional network rewiring observed in different sets of melanoma cell lines using whole genome gene expression microarray profiles. We assess evidence for network rewiring in melanoma patient tumor samples using RNA-sequence data available from The Cancer Genome Atlas. We make a distinction between "unsupervised" and "supervised" network-based methods and contrast their use in identifying consistent differences in networks between subsets of cell lines and tumor samples. We find that different genes play more central roles within subsets of genes within a broader network and hence are likely to be better drug targets in a disease state. Ultimately, we argue that our results have important implications for understanding the molecular pathology of melanoma as well as the choice of treatments to combat that pathology.

Nonlinear mixed effects dose response modeling in high throughput drug screens: application to melanoma cell line analysis.

Cancer cell lines are often used in high throughput drug screens (HTS) to explore the relationship between cell line characteristics and responsiveness to different therapies. Many current analysis methods infer relationships by focusing on one aspect of cell line drug-specific dose-response curves (DRCs), the concentration causing 50% inhibition of a phenotypic endpoint (IC50). Such methods may overlook DRC features and do not simultaneously leverage information about drug response patterns across cell lines, potentially increasing false positive and negative rates in drug response associations. We consider the application of two methods, each rooted in nonlinear mixed effects (NLME) models, that test the relationship relationships between estimated cell line DRCs and factors that might mitigate response. Both methods leverage estimation and testing techniques that consider the simultaneous analysis of different cell lines to draw inferences about any one cell line. One of the methods is designed to provide an omnibus test of the differences between cell line DRCs that is not focused on any one aspect of the DRC (such as the IC50 value). We simulated different settings and compared the different methods on the simulated data. We also compared the proposed methods against traditional IC50-based methods using 40 melanoma cell lines whose transcriptomes, proteomes, and, importantly, BRAF and related mutation profiles were available. Ultimately, we find that the NLME-based methods are more robust, powerful and, for the omnibus test, more flexible, than traditional methods. Their application to the melanoma cell lines reveals insights into factors that may be clinically useful.

Analysis of variability in high throughput screening data: applications to melanoma cell lines and drug responses.

High-throughput screening (HTS) strategies and protocols have undergone significant development in the last decade. It is now possible to screen hundreds of thousands of compounds, each exploring multiple biological phenotypes and parameters, against various cell lines or model systems in a single setting. However, given the vast amount of data such studies generate, the fact that they use multiple reagents, and are often technician-intensive, questions have been raised about the variability, reliability and reproducibility of HTS results. Assessments of the impact of the multiple factors in HTS studies could arguably lead to more compelling insights into the robustness of the results of a particular screen, as well as the overall quality of the study. We leveraged classical, yet highly flexible, analysis of variance (ANOVA)-based linear models to explore how different factors contribute to the variation observed in a screening study of four different melanoma cell lines and 120 drugs over nine dosages studied in two independent academic laboratories. We find that factors such as plate effects, appropriate dosing ranges, and to a lesser extent, the laboratory performing the screen, are significant predictors of variation in drug responses across the cell lines. Further, we show that when sources of variation are quantified and controlled for, they contextualize claims of inconsistencies and reveal the overall quality of the HTS studies performed at each participating laboratory. In the context of the broader screening study, we show that our analysis can also elucidate the robust effects of drugs, even those within specific cell lines.