The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster.

Here we present a statistically rigorous approach to quantifying microarray expression data that allows the relative effects of multiple classes of treatment to be compared and incorporates analytical methods that are common to quantitative genetics. From the magnitude of gene effects and contributions of variance components, we find that gene expression in adult flies is affected most strongly by sex, less so by genotype and only weakly by age (for 1- and 6-wk flies); in addition, sex x genotype interactions may be present for as much as 10% of the Drosophila transcriptome. This interpretation is compromised to some extent by statistical issues relating to power and experimental design. Nevertheless, we show that changes in expression as small as 1.2-fold can be highly significant. Genotypic contributions to transcriptional variance may be of a similar magnitude to those relating to some quantitative phenotypes and should be considered when assessing the significance of experimental treatments.

Assessment of variability in GWAS with CRLMM genotyping algorithm on WTCCC coronary artery disease.

The robustness of genome-wide association study (GWAS) results depends on the genotyping algorithms used to establish the association. This paper initiated the assessment of the impact of the Corrected Robust Linear Model with Maximum Likelihood Classification (CRLMM) genotyping quality on identifying real significant genes in a GWAS with large sample sizes. With microarray image data from the Wellcome Trust Case-Control Consortium (WTCCC), 1991 individuals with coronary artery disease (CAD) and 1500 controls, genetic associations were evaluated under various batch sizes and compositions. Experimental designs included different batch sizes of 250, 350, 500, 2000 samples with different distributions of cases and controls in each batch with either randomized or simply combined (4:3 case-control ratios) or separate case-control samples as well as whole 3491 samples. The separate composition could create 2-3% discordance in the single nucleotide polymorphism (SNP) results for quality control/statistical analysis and might contribute to the lack of reproducibility between GWAS. CRLMM shows high genotyping accuracy and stability to batch effects. According to the genotypic and allelic tests (P<5.0 x 10(-7)), nine significant signals on chromosome 9 were found consistently in all batch sizes with combined design. Our findings are critical to optimize the reproducibility of GWAS and confirm the genetic role in the pathophysiology of CAD.

Comparison of performance of one-color and two-color gene-expression analyses in predicting clinical endpoints of neuroblastoma patients.

Microarray-based prediction of clinical endpoints may be performed using either a one-color approach reflecting mRNA abundance in absolute intensity values or a two-color approach yielding ratios of fluorescent intensities. In this study, as part of the MAQC-II project, we systematically compared the classification performance resulting from one- and two-color gene-expression profiles of 478 neuroblastoma samples. In total, 196 classification models were applied to these measurements to predict four clinical endpoints, and classification performances were compared in terms of accuracy, area under the curve, Matthews correlation coefficient and root mean-squared error. Whereas prediction performance varied with distinct clinical endpoints and classification models, equivalent performance metrics were observed for one- and two-color measurements in both internal and external validation. Furthermore, overlap of selected signature genes correlated inversely with endpoint prediction difficulty. In summary, our data strongly substantiate that the choice of platform is not a primary factor for successful gene expression based-prediction of clinical endpoints.

Assessing sources of inconsistencies in genotypes and their effects on genome-wide association studies with HapMap samples.

The discordance in results of independent genome-wide association studies (GWAS) indicates the potential for Type I and Type II errors. We assessed the repeatibility of current Affymetrix technologies that support GWAS. Reasonable reproducibility was observed for both raw intensity and the genotypes/copy number variants. We also assessed consistencies between different SNP arrays and between genotype calling algorithms. We observed that the inconsistency in genotypes was generally small at the specimen level. To further examine whether the differences from genotyping and genotype calling are possible sources of variation in GWAS results, an association analysis was applied to compare the associated SNPs. We observed that the inconsistency in genotypes not only propagated to the association analysis, but was amplified in the associated SNPs. Our studies show that inconsistencies between SNP arrays and between genotype calling algorithms are potential sources for the lack of reproducibility in GWAS results.

Genomic indicators in the blood predict drug-induced liver injury.

Genomic biomarkers for the detection of drug-induced liver injury (DILI) from blood are urgently needed for monitoring drug safety. We used a unique data set as part of the Food and Drug Administration led MicroArray Quality Control Phase-II (MAQC-II) project consisting of gene expression data from the two tissues (blood and liver) to test cross-tissue predictability of genomic indicators to a form of chemically induced liver injury. We then use the genomic indicators from the blood as biomarkers for prediction of acetaminophen-induced liver injury and show that the cross-tissue predictability of a response to the pharmaceutical agent (accuracy as high as 92.1%) is better than, or at least comparable to, that of non-therapeutic compounds. We provide a database of gene expression for the highly informative predictors, which brings biological context to the possible mechanisms involved in DILI. Pathway-based predictors were associated with inflammation, angiogenesis, Toll-like receptor signaling, apoptosis, and mitochondrial damage. The results show for the first time and support the hypothesis that genomic indicators in the blood can serve as potential diagnostic biomarkers predictive of DILI.

Prostate-specific antigen doubling times are similar in patients with recurrence after radical prostatectomy or radiotherapy: a novel analysis.

PURPOSE: Some investigators have analyzed the rate of growth of prostate cancer that has recurred after definitive radiotherapy or radical prostatectomy using serum prostate-specific antigen (PSA) doubling times (DT). We examined all PSA values in recurrent patients to determine the pattern and rate of increase in PSA after radiation therapy and radical prostatectomy. PATIENTS AND METHODS: Charts of 96 recurrent radical prostatectomy patients (mean age, 62.8 years; range, 47 to 76) and 42 recurrent radiation therapy patients (mean age, 67.2 years; range, 52 to 83) were reviewed. All available PSA values between the date of operation/radiation treatment and last follow-up evaluation or the initiation of second-line therapy are included. Rate of PSA DT was not assumed to be constant over time; it was instead allowed to vary. We use a piecewise linear random-coefficients model in time for log (PSA), which allowed different mean models for both treatments. RESULTS: The PSA DT in the first year after radiation therapy was--1.17 years, which reflects the continuous decline in PSA in the average patients during the first year after radiotherapy despite eventual biochemical progression. In contrast, the PSA DT in the radical prostatectomy group was 0.66 in the first year. In year 2, after radiation therapy, the PSA DT was lengthy at 1.82 years, significantly longer (P = .0025) than in the radical prostatectomy group (0.76 years). After year 2, there were no significant differences between the two groups (P > .05). CONCLUSION: A piecewise linear random-coefficients model enables interval analysis of PSA DT. While the PSA DT after radiation therapy and radical prostatectomy are different in the first 2 years, the rate of increase in PSA appears to be similar in the two groups after year 2, which suggests the rate of growth of cancers that recur after radiation therapy and radical prostatectomy is similar.

Assessing gene significance from cDNA microarray expression data via mixed models.

The determination of a list of differentially expressed genes is a basic objective in many cDNA microarray experiments. We present a statistical approach that allows direct control over the percentage of false positives in such a list and, under certain reasonable assumptions, improves on existing methods with respect to the percentage of false negatives. The method accommodates a wide variety of experimental designs and can simultaneously assess significant differences between multiple types of biological samples. Two interconnected mixed linear models are central to the method and provide a flexible means to properly account for variability both across and within genes. The mixed model also provides a convenient framework for evaluating the statistical power of any particular experimental design and thus enables a researcher to a priori select an appropriate number of replicates. We also suggest some basic graphics for visualizing lists of significant genes. Analyses of published experiments studying human cancer and yeast cells illustrate the results.

An example of using mixed models and PROC MIXED for longitudinal data.

Longitudinal data, or data that are repeated measurements on various subjects across time, are commonplace in biostatistical studies. The general linear mixed model is a useful statistical tool for analyzing such data and drawing meaningful inferences about them. This paper discusses some of the most common mixed models and fits them to a prototypical example involving repeated measures on blood pressure. Computer implementation is via the MIXED procedure in the SAS System, and code descriptions and output interpretations accompany the example.

Nonconjugate Bayesian analysis of variance component models.

We consider the usual normal linear mixed model for variance components from a Bayesian viewpoint. With conjugate priors and balanced data, Gibbs sampling is easy to implement; however, simulating from full conditionals can become difficult for the analysis of unbalanced data with possibly nonconjugate priors, thus leading one to consider alternative Markov chain Monte Carlo schemes. We propose and investigate a method for posterior simulation based on an independence chain. The method is customized to exploit the structure of the variance component model, and it works with arbitrary prior distributions. As a default reference prior, we use a version of Jeffreys' prior based on the integrated (restricted) likelihood. We demonstrate the ease of application and flexibility of this approach in familiar settings involving both balanced and unbalanced data.

Analysis of data from group-randomized trials with repeat observations on the same groups.

This study used Monte Carlo simulations to evaluate the performance of alternative models for the analysis of group-randomized trials having more than two time intervals for data collection. The major distinction among the models tested was the sampling variance of the intervention effect. In the mixed-model ANOVA, the sampling variance of the intervention effect is based on the variance among group x time-interval means. In the random coefficients model, the sampling variance of the intervention effect is based on the variance among the group-specific slopes. These models are equivalent when the design includes only two time intervals, but not when there are more than two time intervals. The results indicate that the mixed-model ANOVA yields unbiased estimates of sampling variation and nominal type I error rates when the group-specific time trends are homogenous. However, when the group-specific time trends are heterogeneous, the mixed-model ANOVA yields downwardly biased estimates of sampling variance and inflated type I error rates. In contrast, the random coefficients model yields unbiased estimates of sampling variance and the nominal type I error rate regardless of the pattern among the groups. We discuss implications for the analysis of group-randomized trials with more than two time intervals.