Motif and consensus sequence matching was performed using the Transcription Ele

Many efforts have been made in search of common molecular signatures, however without obvious success. This is partly due to the highly heterogeneous nature of cancer. Tumour samples often comprise of subpopula tions with different genomic alterations. However, [You must be registered and logged in to see this link.] the most popular outlier detection algorithm, t test or its analogues, simply removes heterogeneity between sub types, and fail to identify the subgroup specific gene alterations. Recently novel statistical methods were developed to identify patterns only existed in the sub groups of the studied samples. In this study, we applied these outlier detection meth ods to analyze our collection of four miRNA expression microarray datasets to identify differentially expressed miRNAs. The DE miRNAs were then compared among the four data sets at both gene and gene set levels for comparison.

By considering the cancer heterogeneity, we applied different statistical methods to identify the consistent prostate cancer associated [You must be registered and logged in to see this link.] pathways that are coordinately targeted by miRNAs. Results Comparison of heterogeneous feature detection algorithms Most of the previous expression data studies used fold change, t test and other statistics alike to detect cancer related genes. Recently, it has been recognized that many oncogenes show altered expression in only a small proportion of cancer samples. Such features will be removed when using t test or t test like methods because they average gene expression levels in all the studied samples. Tomlins et al. concluded that t tests were not adequate for detecting heterogeneous patterns of oncogenes.

[You must be registered and logged in to see this link.] To address this complexity, a series of new heteroge neous detection algorithms have been proposed in recent years. Among these methods are Least Sum of Ordered Subset Squared, Cancer Outlier Profile Analysis, Maximum Ordered Subset T statistics, Outlier Robust T statistics, and Outlier Sum. The performance of the above algorithms and the traditional t test were compared on the detection of the outliers in our collection of prostate cancer associated microRNA expression data. The outliers here refer to the deferentially expressed microRNAs. For all these methods applied to the dif ferent data sets with different numbers of samples, we set the quantile of outliers to 0. 05.

Those DE miRNAs detected by at least three methods were con sidered to be putative PCa associated outliers, and then the percentages of the putative outliers in the original result of each method were calculated to determine the methods accuracy. In most of the cases, these heterogeneity feature detection algorithms per formed better than the traditional t test. In most of this comparison, ORT performed better than the other methods. For these four studied datasets, ORT had the biggest median observation and smallest standard deviation. Therefore, we take the result by ORT for the downstream analyses. The outlier miRNA targets in prostate cancer As miRNAs play a role in post transcriptional regulation by targeting complementary mRNAs, we collection their putative targets and subsequently mapped these target genes to pathways or gene sets for enrichment analysis. Target genes were retrieved from both TargetScan data base and our integrative prediction.