The above estimators were used for generating random realization of univariate data for respective conditions. The steps involved in the analysis are given as below: Step1: Generate a simulated

dataset using the estimated parameters from Equations (1) and (2) for all genes. Obtain moderated t-statistic values for the simulated dataset. Similarly, simulate 100 datasets and obtain moderated t-statistic values for the respective simulated dataset. Gene expression profiles of 89 Homo sapiens prostate samples were downloaded from a publicly available Raf inhibitor database, ArrayExpress, of which, 34 were African–American prostate tumor samples, 35 were European–American prostate tumor samples, and 20 were cancer-free samples. In the present study, our interest was to compare 35 European–American with 34 African–American patients to detect the true significant genes that are involved in the prostate cancer progression. In literature, there are many sophisticated analytical and statistical approaches that were proposed to microarray normalization and differential

expression analysis. Selleck SCR7 In the present analysis, the data was log transformed and normalized with median centering. The median absolute deviation was also performed on the datasets for uniformity of scale. The moderated t-statistic was applied on normalized dataset and for each simulated dataset, to detect true significant genes (see methods). The sorted observed

t-statistic values from normalized data and the sorted expected t- statistic values from simulated datasets are shown in Fig. 1. The set of significant genes identified at different thresholds (δ0) are given in Table 1. We obtained MDS classification of both tumor-groups of 34 African–American and 35 European–American samples (patients) over from each set of significant genes and correspondingly from the subset of significant genes. The classification of both tumor-groups was poor from all set of significant genes. The number of correctly classified and misclassified samples is also shown in Table 1. The samples GSE6956GSM160352, GSE6956GSM160358, GSE6956GSM160378 from African–American prostate tumors and GSE6956GSM160416, GSE6956GSM160379, and GSE6956GSM160365 from European–American prostate tumors were often misclassified. Hence, all these samples were eliminated from analysis and continued the analysis from step 1 to step 5 as mentioned in the methods. By excluding the above 6 samples, new moderated t-statistic values were obtained on normalized data and correspondingly for simulated datasets. The number of significant genes identified by choosing different thresholds is shown in Table 2. At a threshold of δ0 = 0.