Here, we consider a Bayesian model based on gene frequency (BGF) where both conditional means and variances are modeled as a function of the conditional gene frequencies at the QTL.
The reason why we did not find a separate ERBB2-cluster might be related to the fact that the intrinsic gene list is based solely on breast tumors, while our clustering is based on genes that show variance between all types of epithelial tumors, or in other words our gene list is both intrinsic and extrinsic.
Classifiers based on genes had a very large variance, compared to the BC module-based classifier.
This was true for the selected set of genes based on high variance genes but also visible using the complete transcriptome.
The independent t test for each experiment was able to detect genes that were significantly expressed even with low expression ratios because unknown error variance components were estimated based on a gene-specific variance for each gene.
The top 200 genes based on their variance across samples or genes that were obtained by statistical filtering were subjected to hierarchical clustering analysis (Manhattan distance, average metric).
It is necessary to point out that the statistical model that is based on a common gene variance assumption has more power simply because it assumes that all genes in the genome have the same variation; which is unlikely to be true across nearly 26,000 genes.
We examined the rate of evolution between specific classes of genes, based on the variance of the log2 FC distributions for the genes and found significant differences in the rate of evolution between groups of genes.
Processed probe data were then filtered according to the following criteria: minimal signal intensity fold-change of 1.50 across all samples; minimal absolute change in probe signal intensity of 150 across all samples, based on the variance of gene expression, with a 10% threshold.
Secondly, in concordance with WGCNA, which selects genes based on their variance, the most varying SNPs in the population were selected due to computational limitations.
