Bootstrapping cluster analysis: assessing the reliability of conclusions from microarray experiments.

We introduce a general technique for making statistical inference from clustering tools applied to gene expression microarray data. The approach utilizes an analysis of variance model to achieve normalization and estimate differential expression of genes across multiple conditions. Statistical inference is based on the application of a randomization technique, bootstrapping. Bootstrapping has previously been used to obtain confidence intervals for estimates of differential expression for individual genes. Here we apply bootstrapping to assess the stability of results from a cluster analysis. We illustrate the technique with a publicly available data set and draw conclusions about the reliability of clustering results in light of variation in the data. The bootstrapping procedure relies on experimental replication. We discuss the implications of replication and good design in microarray experiments.

Statistical design and the analysis of gene expression microarray data.

Gene expression microarrays are an innovative technology with enormous promise to help geneticists explore and understand the genome. Although the potential of this technology has been clearly demonstrated, many important and interesting statistical questions persist. We relate certain features of microarrays to other kinds of experimental data and argue that classical statistical techniques are appropriate and useful. We advocate greater attention to experimental design issues and a more prominent role for the ideas of statistical inference in microarray studies.

Experimental design for gene expression microarrays.

We examine experimental design issues arising with gene expression microarray technology. Microarray experiments have multiple sources of variation, and experimental plans should ensure that effects of interest are not confounded with ancillary effects. A commonly used design is shown to violate this principle and to be generally inefficient. We explore the connection between microarray designs and classical block design and use a family of ANOVA models as a guide to choosing a design. We combine principles of good design and A-optimality to give a general set of recommendations for design with microarrays. These recommendations are illustrated in detail for one kind of experimental objective, where we also give the results of a computer search for good designs.

Analysis of variance for gene expression microarray data.

Spotted cDNA microarrays are emerging as a powerful and cost-effective tool for large-scale analysis of gene expression. Microarrays can be used to measure the relative quantities of specific mRNAs in two or more tissue samples for thousands of genes simultaneously. While the power of this technology has been recognized, many open questions remain about appropriate analysis of microarray data. One question is how to make valid estimates of the relative expression for genes that are not biased by ancillary sources of variation. Recognizing that there is inherent "noise" in microarray data, how does one estimate the error variation associated with an estimated change in expression, i.e., how does one construct the error bars? We demonstrate that ANOVA methods can be used to normalize microarray data and provide estimates of changes in gene expression that are corrected for potential confounding effects. This approach establishes a framework for the general analysis and interpretation of microarray data.

Do pulse checks cause a significant delay in the initial defibrillation sequence?

This study was undertaken to determine if checking for a pulse between initial defibrillations causes a clinically significant delay in the administration of the defibrillations. Ten emergency department nurses and 10 emergency medicine resident physicians were timed delivering three successive defibrillations (200, 300 and 360 J) to a manikin under three randomly assigned scenarios: (1) without pulse checks; (2) with pulse checks performed by an assistant; and (3) with pulse checks performed by the participant. All participants performed the three defibrillation scenarios using three different models of defibrillators. Repeated measures analysis of variance was used to compare mean defibrillation times for the three scenarios. The mean time was 20.4 +/- 1.0 s for defibrillation without pulse checks; 20.2 +/- 1.2 s with pulse checks by an assistant and 22.0 +/- 2.0 s with pulse checks by the participant. There was a statistically significant difference between no pulse checks and pulse checks by the participant. No statistically significant difference was noted between no pulse checks and pulse checks by an assistant. We conclude that checking for a pulse does cause a statistically significant delay in the administration of defibrillations. This difference, however, is not likely to be clinically relevant.