Finding known co-expressed genes that are not co-regulated in TCGA level 3 data?

Question

[kbun]

I have the expressions of a set of co-expressed genes (from a different paper) across 30 normal (N) and tumor (T) tissues of different cancer types (normalized TCGA Level 3 Data) in a matrix.

Not all cancers have (N) expression data, so there are 10 (N) samples and about 20 (T) samples. Each (N) sample has a matching (T) sample, so there are 10 (T) samples that don't have a matching normal.

How can I find out statistically if any of these co-expressed genes are not co-regulated in any Cancer Tumor tissue?

Manual Example: gene1 and gene2 have been previously found to be co-expressed in normal tissue in a prior paper. In Breast Cancer, it is found that gene1 was upregulated from 10 to 50,000 rpm, but gene2 was downregulated from 500 to 10 rpm (actual data is going to be far less extreme, of course)

I was planning on constructing a gene co-expression network and have an output of genes that aren't significantly 'co-regulated', but I don't think this is a statistically sound approach.

Any help would be great! Thanks!

Answer 1

[Khader Shameer]

You can use a reference database to compile a background distribution of canonical co-expressed genes

• STRING-db (filter by organism and interaction set to "co-expression" and/or "gene co-occurrence" depending on your preference
• GeneMania (filter by co-expression)