Clinical trial simulation to assist in COPD trial planning and design with a biomarker-based diagnostic: when to pull the trigger?

Chronic Obstructive Pulmonary Disease (COPD) is a heterogeneous disease with a wide range of clinical phenotypes that vary from predominantly airway disease (chronic bronchitis) to predominantly parenchymal disease (emphysema). Current advances for the treatment of COPD are increasingly focused on targeted treatments and development of novel biomarker-based diagnostics (Dx)'s to select the patients most likely to benefit. Clinical trial planning and design with biomarkers includes additional considerations beyond those for conventional trials in un-selected populations, e.g., the heterogeneity of COPD phenotypes in the population, the ability of a biomarker to predict clinically meaningful phenotypes that are differentially associated with the response to a targeted treatment, and the data needed to make Go/No Go decisions during clinical development. We developed the Clinical Trial Object Oriented Research Application (CTOORA), a computer-aided clinical trial simulator of COPD patient outcomes, to inform COPD trial planning with biomarkers. CTOORA provides serial projections of trial success for a range of hypothetical and plausible scenarios of interest. In the absence of data, CTOORA can identify characteristics of a biomarker-based Dx needed to provide a meaningful advantage when used in a clinical trial. We present a case study in which CTOORA is used to identify the scenarios for which a biomarker may be used successfully in clinical development. CTOORA is a tool for robust clinical trial planning with biomarkers, to guide early-to-late stage development that is positioned for success.

Error control variability in pathway-based microarray analysis.

MOTIVATION: The decision to commit some or many false positives in practice rests with the investigator. Unfortunately, not all error control procedures perform the same. Our problem is to choose an error control procedure to determine a P-value threshold for identifying differentially expressed pathways in high-throughput gene expression studies. Pathway analysis involves fewer tests than differential gene expression analysis, on the order of a few hundred. We discuss and compare methods for error control for pathway analysis with gene expression data. RESULTS: In consideration of the variability in test results, we find that the widely used Benjamini and Hochberg's (BH) false discovery rate (FDR) analysis is less robust than alternative procedures. BH's error control requires a large number of hypothesis tests, a reasonable assumption for differential gene expression analysis, though not the case with pathway-based analysis. Therefore, we advocate through a series of simulations and applications to real gene expression data that researchers control the number of false positives rather than the FDR.

Discovery of the Selective CYP17A1 Lyase Inhibitor BMS-351 for the Treatment of Prostate Cancer.

Efforts to identify a potent, reversible, nonsteroidal CYP17A1 lyase inhibitor with good selectivity over CYP17A1 hydroxylase and CYPs 11B1 and 21A2 for the treatment of castration-resistant prostate cancer (CRPC) culminated in the discovery of BMS-351 (compound 18), a pyridyl biaryl benzimidazole with an excellent in vivo profile. Biological evaluation of BMS-351 at a dose of 1.5 mg in castrated cynomolgus monkeys revealed a remarkable reduction in testosterone levels with minimal glucocorticoid and mineralcorticoid perturbation. Based on a favorable profile, BMS-351 was selected as a candidate for further preclinical evaluation.

A comprehensive assessment of p53-responsive genes following adenoviral-p53 gene transfer in Bcl-2-expressing prostate cancer cells.

The p53 protein can induce cell cycle arrest or apoptosis following activation in response to DNA damage. The function of p53 is largely mediated by regulating the expression of downstream target genes. Adenoviral-p53 gene transfer (Ad-p53) is currently being evaluated in clinical trials as a therapeutic intervention. Tumor response is likely to be influenced by context-dependent variables, such as expression of bcl-2. Bcl-2 is upregulated in a variety of neoplasms, and can inhibit p53-dependent apoptosis. It was therefore of interest to use a global genomic strategy to assess gene expression following Ad-p53 gene transfer and to determine if the expression of specific Ad-p53-responsive genes could be modulated in the context of bcl-2 gene deregulation. cDNA arrays were used to identify p53-responsive genes following Ad-p53 gene transfer in control and bcl-2-overexpressing PC3 prostate cancer cells. A total of 40 transcripts were significantly upregulated by Ad-p53 in both control and bcl-2-transfectant PC3 cells. Conversely, 19 transcripts were significantly repressed in both cell lines. These Ad-p53-responsive transcripts included previously identified p53 targets, known genes representing candidate p53 targets, and transcripts identified as expressed sequence tags. A subset of 15 transcripts was differentially modulated by Ad-p53 in the context of bcl-2. Some of these genes were also differentially modulated in LNCaP (wt p53) cells following DNA damage. These results document a number of potential p53 targets and mediators of therapeutically relevant genotoxic stress. The findings further suggest that bcl-2 may inhibit cell death at multiple points downstream of p53 activation.

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.

Inter-gene correlation on oligonucleotide arrays: how much does normalization matter?

BACKGROUND AND OBJECTIVE: Normalization is a standard low-level preprocessing procedure in microarray data analysis to minimize the systematic technological variations and produce more reliable results. A variety of normalization approaches have been introduced and are widely applied. Normalization, however, remains controversial. The sensitivity of array results to normalization is an open question. No clear standard for comparing or judging normalization methods has yet emerged, and the effects of normalization on gene-to-gene co-expression are unclear. METHODS: In this investigation, we applied 1-, 2-, and N-quantile normalization to several publicly available microarray datasets quantified with either MAS 5.0 or dCHIP and evaluated the effect on gene-to-gene co-expression. We introduced a graphical method to explore trends in gene correlation. RESULTS: We found clear differences in the distributions of gene dependencies by normalization method. Increasing the number of standardized quantiles in the normalization reduced trends in correlation by signal intensity in MAS 5.0 quantifications but not dCHIP. Increasing the number of standardized quantiles did not markedly reduce the correlation of known overlapping targets with MAS 5.0. CONCLUSIONS: Normalization plays a very important role in the estimation of inter-gene dependency. Caution should be used when making inferences concerning gene-wise dependencies with microarrays until this source of variation is better understood.

The Thoc1 ribonucleoprotein and prostate cancer progression.

BACKGROUND: The majority of newly diagnosed prostate cancers will remain indolent, but distinguishing between aggressive and indolent disease is imprecise. This has led to the important clinical problem of overtreatment. THOC1 encodes a nuclear ribonucleoprotein whose expression is higher in some cancers than in normal tissue. The hypothesis that THOC1 may be a functionally relevant biomarker that can improve the identification of aggressive prostate cancer has not been tested. METHODS: THOC1 protein immunostaining was evaluated in a retrospective collection of more than 700 human prostate cancer specimens and the results associated with clinical variables and outcome. Thoc1 was conditionally deleted in an autochthonous mouse model (n = 22 or 23 per genotype) to test whether it is required for prostate cancer progression. All statistical tests were two-sided. RESULTS: THOC1 protein immunostaining increases with higher Gleason score and more advanced Tumor/Node/Metastasis stage. Time to biochemical recurrence is statistically significantly shorter for cancers with high THOC1 protein (log-rank P = .002, and it remains statistically significantly associated with biochemical recurrence after adjusting for Gleason score, clinical stage, and prostate-specific antigen levels (hazard ratio = 1.61, 95% confidence interval = 1.03 to 2.51, P = .04). Thoc1 deletion prevents prostate cancer progression in mice, but has little effect on normal tissue. Prostate cancer cells deprived of Thoc1 show gene expression defects that compromise cell growth. CONCLUSIONS: Thoc1 is required to support the unique gene expression requirements of aggressive prostate cancer in mice. In humans, high THOC1 protein immunostaining associates with prostate cancer aggressiveness and recurrence. Thus, THOC1 protein is a functionally relevant molecular marker that may improve the identification of aggressive prostate cancers, potentially reducing overtreatment.

Gene silencing in cancer by histone H3 lysine 27 trimethylation independent of promoter DNA methylation.

Epigenetic silencing in cancer cells is mediated by at least two distinct histone modifications, polycomb-based histone H3 lysine 27 trimethylation (H3K27triM) and H3K9 dimethylation. The relationship between DNA hypermethylation and these histone modifications is not completely understood. Using chromatin immunoprecipitation microarrays (ChIP-chip) in prostate cancer cells compared to normal prostate, we found that up to 5% of promoters (16% CpG islands and 84% non-CpG islands) were enriched with H3K27triM. These genes were silenced specifically in prostate cancer, and those CpG islands affected showed low levels of DNA methylation. Downregulation of the EZH2 histone methyltransferase restored expression of the H3K27triM target genes alone or in synergy with histone deacetylase inhibition, without affecting promoter DNA methylation, and with no effect on the expression of genes silenced by DNA hypermethylation. These data establish EZH2-mediated H3K27triM as a mechanism of tumor-suppressor gene silencing in cancer that is potentially independent of promoter DNA methylation.

Enrichment analysis in high-throughput genomics - accounting for dependency in the NULL.

Translating the overwhelming amount of data generated in high-throughput genomics experiments into biologically meaningful evidence, which may for example point to a series of biomarkers or hint at a relevant pathway, is a matter of great interest in bioinformatics these days. Genes showing similar experimental profiles, it is hypothesized, share biological mechanisms that if understood could provide clues to the molecular processes leading to pathological events. It is the topic of further study to learn if or how a priori information about the known genes may serve to explain coexpression. One popular method of knowledge discovery in high-throughput genomics experiments, enrichment analysis (EA), seeks to infer if an interesting collection of genes is 'enriched' for a Consortium particular set of a priori Gene Ontology Consortium (GO) classes. For the purposes of statistical testing, the conventional methods offered in EA software implicitly assume independence between the GO classes. Genes may be annotated for more than one biological classification, and therefore the resulting test statistics of enrichment between GO classes can be highly dependent if the overlapping gene sets are relatively large. There is a need to formally determine if conventional EA results are robust to the independence assumption. We derive the exact null distribution for testing enrichment of GO classes by relaxing the independence assumption using well-known statistical theory. In applications with publicly available data sets, our test results are similar to the conventional approach which assumes independence. We argue that the independence assumption is not detrimental.

Identification of cell cycle regulatory genes as principal targets of p53-mediated transcriptional repression.

Historically, most studies attribute p53 function to the transactivation of target genes. That p53 can selectively repress genes to affect a cellular response is less widely appreciated. Available evidence suggests that repression is important for p53-induced apoptosis and cell cycle arrest. To better establish the scope of p53-repressed target genes and the cellular processes they may affect, a global expression profiling strategy was used to identify p53-responsive genes following adenoviral p53 gene transfer (Ad-p53) in PC3 prostate cancer cells. A total of 111 genes, 0.77% of the 14,500 genes represented on the Affymetrix U133A microarray, were repressed more than 2-fold (p < or = 0.05). Validation of the array data, using reverse transcription-PCR of 20 randomly selected genes, yielded a confirmation rate of >95.5% for the complete data set. Functional over-representation analysis revealed that cell cycle regulatory genes exhibited a highly significant enrichment (p < or = 5 x 10(-28)) within the transrepressed targets. 41% of the repressed targets are cell cycle regulators. A subset of these genes exhibited repression following DNA damage, preceding cell cycle arrest, in LNCaP cells. The use of a p53 small interfering RNA strategy in LNCaP cells and the use of p53-null cell lines demonstrated that this repression is p53-dependent. These findings identify a set of genes not known previously to be down-regulated by p53 and indicate that p53-induced cell cycle arrest is a function of not only the transactivation of cell cycle inhibitors (e.g. p21) but also the repression of targets that regulate proliferation at several distinct phases of the cell cycle.

Microtubule-associated protein tau: a marker of paclitaxel sensitivity in breast cancer.

Breast cancers show variable sensitivity to paclitaxel. There is no diagnostic test to identify tumors that are sensitive to this drug. We used U133A chips to identify genes that are associated with pathologic complete response (pCR) to preoperative paclitaxel-containing chemotherapy in stage I-III breast cancer (n = 82). Tau was the most differentially expressed gene. Tumors with pCR had significantly lower (P < 0.3 x 10(-5)) mRNA expression. Tissue arrays from 122 independent but similarly treated patients were used for validation by immunohistochemistry. Seventy-four percent of pCR cases were tau protein negative; the odds ratio for pCR was 3.7 (95% confidence interval, 1.6-8.6; P = 0.0013). In multivariate analysis, nuclear grade (P < 0.01), age <50 (P = 0.03), and tau-negative status (P = 0.04) were independent predictors of pCR. Small interfering RNA experiments were performed to examine whether down-regulation of tau increases sensitivity to chemotherapy in vitro. Down-regulation of tau increased sensitivity of breast cancer cells to paclitaxel but not to epirubicin. Tubulin polymerization assay was used to assess whether tau modulates binding of paclitaxel to tubulin. Preincubation of tubulin with tau resulted in decreased paclitaxel binding and reduced paclitaxel-induced microtubule polymerization. These data suggest that low tau expression renders microtubules more vulnerable to paclitaxel and makes breast cancer cells hypersensitive to this drug. Low tau expression may be used as a marker to select patients for paclitaxel therapy. Inhibition of tau function might be exploited as a therapeutic strategy to increase sensitivity to paclitaxel.