Quantification of Pathological Bias in the SignalomeLlevel Analyzing the Transcriptome-Wide Data

Abstract

Investigation of huge sets of transcriptomic data is commonly used now for gene expression profiling, e.g., of cancer patients with different tumor localization and morphology. Recently, we have proposes a method for aggregation of gene expression data into special subsets that may be used for characterization and quantification of pathological changes for individual patients at the level of virtually every known cell signaling pathway. The quantitative measure of relative pathological changes, whether it is over-activation or inhibition, of each signaling pathway, which we have suggested in our software packages OncoFinder] and GeroScope, accumulates the transcriotomic data according to the relative roles of different gene products in activation of signaling pathways.

Here, we test the abilities of our method to detect the meaningful information on the overall pattern of pathological changes for patients with malignant tumors. It is not a secret that the transcriptomic data does not always correlate well even for the same samples that were investigated using different methods and platforms. We gave performed a comparison of correlation coefficients between the data obtained using a RNA microarraying and RNA sequencing devices for three datasets of malignant cells. The results obtained shows that the aggregation of the transcriptomic data on the pool of genes into “higher-level” sets that correspond to distinct signaling pathways can “restore” the correlation between the mirroarray and RNA sequencing data, thus “extracting” the meaningful signal from the transcriptome-wide “noise”.

We also apply our method for pathway perturbation analysis to the search for markers of cancer types with different morphology and localization. The data pool were obtained the pool of data on gene expression levels on different cancer nosologic forms partly using the results our of own investigations of samples taken from the bladder carcinoma and various head and neck tumors, and partly using the information published in the Gene expression omnibus (GEO) repository. For each of eight nosologic forms that we have investigated using the pathway actiation strength (PAS) concept, we have found the specific pathway markers, as well as discriminators of one cancer type from another.