Search

Typically, in high-throughput data, the number of features is often substantially greater than the number of samples. One approach to this statistical challenge is to perform feature selection and usually, only predictive accuracy is used to perform feature selection. We present a new feature selection method called Wrapper Consistency Analysis that strives to optimize both the predictive accuracy as well as a measure called overlap (or consistency). We believe this analysis provides additional information about the goodness of a selected set of features.

A potential limitation of data from microarray experiments exists when improper control samples are used. In cancer research, comparisons of tumour expression profiles to those from normal samples is challenging due to tissue heterogeneity (mixed cell populations). A specific example exists in a published colon cancer dataset, in which tissue heterogeneity was reported among the normal samples. In this paper, we show how to overcome or avoid the problem of using normal samples that do not derive from the same tissue of origin as the tumour. We advocate an exploratory unsupervised bootstrap analysis that can reveal unexpected and undesired, but strongly supported, clusters of samples that reflect tissue differences instead of tumour versus normal differences. All of the algorithms used in the analysis, including the maximum difference subset algorithm, unsupervised bootstrap analysis, pooled variance t-test for finding differentially expressed genes and the jackknife to reduce false positives, are incorporated into our online Gene Expression Data Analyzer ( http:// bioinformatics.upmc.edu/GE2/GEDA.html ).

We have developed a new approach to the measurement of phylogenetic signal in character state matrices called relative apparent synapomorphy analysis (RASA). RASA provides a deterministic, statistical measure of natural cladistic hierarchy (phylogenetic signal) in character state matrices. The method works by determining whether a measure of the rate of increase of cladistic similarity among pairs of taxa as a function of phenetic similarity is greater than a null equiprobable rate of increase. Our investigation of the utility and limitations of RASA using simulated and bacteriophage T7 data sets indicates that the method has numerous advantages over existing measures of signal. A first advantage is computational efficiency. A second advantage is that RASA employs known methods of statistical inference, providing measurable sensitivity and power. The performance of RASA is examined under various conditions of branching evolution as the number of characters, character states per character, and mutations per branch length are varied. RASA appears to provide an unbiased and reliable measure of phylogenetic signal, and the general approach promises to be useful in the development of new techniques that should increase the rigor and reliability of phylogenetic estimates.