Adding Confidence to Gene Expression Clustering

Abstract

It has been well established that gene expression data contain large amounts of random variation that affects both the analysis and the results of microarray experiments. Typically, microarray data are either tested for differential expression between conditions or grouped on the basis of profiles that are assessed temporally or across genetic or environmental conditions. While testing differential expression relies on levels of certainty to evaluate the relative worth of various analyses, cluster analysis is exploratory in nature and has not had the benefit of any judgment of statistical inference. By using a novel dissimilarity function to ascertain gene expression clusters and conditional randomization of the data space to illuminate distinctions between statistically significant clusters of gene expression patterns, we aim to provide a level of confidence to inferred clusters of gene expression data. We apply both permutation and convex hull approaches for randomization of the data space and show that both methods can provide an effective assessment of gene expression profiles whose coregulation is statistically different from that expected by random chance alone.

The Association Among Gene Expression Responses to Nine Abiotic Stress Treatments in Arabidopsis thaliana

“…, u i9 } sets that are equally likely to occur at each organtime combination. To generate a randomized data set consisting of N of the N 10 possible sets, the permutation procedure implemented by Munneke et al (2005) was used, in which mean gene expression values ( u) were permuted across genes within each of the nine stress treatments. This permutation procedure yields a randomized data set containing N of the N 10 possible { u i0 , u i1 , .…”

Unsupervised Selection of Highly Coexpressed and Noncoexpressed Genes Using a Consensus Clustering Approach

“…data and compute the p-value for each cluster C above. The dataset is randomly resampled (permutation plus convexhull (Munneke et al, 2005)) a number of times (n r ), for each of which the entire process starting from building the AM with k * selected above to the consensus clustering step is done and random-resulting clusters are returned. Subsequently, the procedure estimates the cluster significance for these random clusters and builds up a distribution of cluster significance.…”

“…Random data are generated through resampling (permutation plus convex-hull) (Munneke et al, 2005) synthetic datasets, each with 10 times. The AM histogram, CDF, AUC, and gap curves were built independently for each dataset and then the average ones are made for each data type to have a consensus view (Fig.…”

“…The significance of a cluster is defined as the product of its size times its homogeneity; as the nature of clustering methods, random data can be divided into a number of large-size clusters depending on the input parameter "nCluster"; but these clusters contain arbitrarily objects (genes); thus, to count this event into the definition of cluster significance, besides cluster size we also take the homogeneity into account. This function will run a number of times (nResamplingTimes) on resampling data (permutation plus convex-hull as default) (Munneke et al, 2005) and estimate the distribution of cluster significance. The cutoff p-value will help to remove trivial clusters.…”

“…In the context of microarray analysis it may include gene ontology (Fang et al, 2006;Kustra and Zagdanskey, 2006), gene annotation (Huang and Pan, 2006), gene function (Pan, 2006) etc. ; finally (4) the significance of a cluster and/or the probability that two genes may belong to one cluster or two different clusters is also an issue (Munneke et al, 2005;van der Laan and Pollard, 2003;Zhang and Zhao, 2000).…”