Differential Expression Analysis For Individual Cancer Samples Based On Robust Within-sample Relative Gene Expression Orderings Across Multiple Profiling Platforms

The within-sample relative expression orderings(REOs) of gene pairs are widely stable in a particular type of human normal tissue but disturbed widely in the corresponding disease tissue. Based on this biological phenomenon, we have recently proposed an algorithm named RankComp to detect differentially expressed genes(DEGs) for any disease sample measured by a platform by comparing the REO pattern of the disease sample with the highly stable REOs predetermined in a collection of normal samples measured by the same platform. However, the current RankComp algorithm is based on REOs of gene pairs that are highly stable in a pre-defined percentage(e.g., 99%) of normal samples, which is lack of statistical control and may limit the detection power of DEGs in individual samples. Thus, it is also necessary to evaluate the performance of RankComp when using significantly stable gene pairs, selected with statistical control rather than a pre-defined percentage, in a particular type of normal tissue as the basis for the individual-level differential expression analysis. In addition, when applying RankComp algorithm to data from different gene expression profiling platforms with different probe design principles, the REOs may be subject to a certain degree of uncertainty. Thus, it is necessary to further evaluate the cross-platform properties of REOs.With 461 normal lung and 234 normal colorectal tissue samples separately measured by four commonly used platforms(Affymetrix, Illumina, Agilent microarray platforms and a RNA_seq platform), we demonstrated that tens of millions of gene pairs with significantly stable REOs can be consistently detected in samples measured by different platforms and using only a small number of samples(about 20 samples) can identify about 80% of the significantly stable REOs for a particular type of human normal tissue. Taking the lung tissue samples for example, based on the significantly stable REOs(FDR<0.01) consistently detected by the four platforms for normal lung tissue, RankComp detected averagely 1184, 1335 and 1116 DEGs per sample with averagely 96.51%, 95.95% and 94.78% precisions in three evaluation datasets with 25, 57 and 58 paired lung cancer and adjacent normal samples, respectively. Individualized pathway analysis revealed some common functional mechanisms and subtype-specific functional mechanisms of lung cancer. Similar results were observed for colorectal cancer.For a particular type of human normal tissue, REOs of gene pairs are widely stable and most of them could be found with only about 20 samples. In addition, significantly stable within-sample REOs especially for gene pairs with large expression differences are largely consistent across samples measured by different platforms. Based on the cross-platform significantly stable REOs for a particular normal tissue, differentially expressed genes and pathways in any disease sample measured by any of the platforms can be readily and accurately detected, which could be further exploited for dissecting the heterogeneity of cancer.