Enrichment Analysis Without Differentially Expressed Protein List

Question

[Woa]

I've some a list of protein Ids derived by proteomics experiment and for two samples, Normal and Cancer cells. However NO quantitative data,relative or absolute is available.

Can I upload these Ids to some functional enrichment analysis tool like GO term over representation analyzer to have a ROUGH Idea of what functional categories are enriched /non-enriched in cancer cell lines compared to Normal. Say, the p-value of cell-division process is 0.05 for normal tissue data and that of cancer is 0.0001, so I can tell that Cell-division proteins are enriched in cancer cells.

How bad theoretically is this method when I don't have any list of differentially expressed proteins both up and down regulated?

[Arun]

What does your proteomics experiment experiment test for or do? Or what is the design of your experiment? What are you looking for? Just identifying proteins? or looking for specific post-translational modifications between samples? etc...

[Woa]

The experiment was mainly done for identifying proteins and for optimization of upstream methods, I wonder if functional enrichment analysis can be done with it.

[Arun]

I think you have more chances of not finding any. 1) your gene set is not very specific 2) there is no testable hypothesis from which your "candidate genes" are derived 3) It is not actually enrichment you're looking for. You are basically looking if there are genes that are totally not expressed in one condition than the other and they may be too small an event to show a pattern of enrichment and might hide in the background.

Actually I am not sure if you find any enrichment, whether they can be misleading (as I mentioned before). They are from proteomic experiments and if you do find an enrichment, they could give you a rough idea... (even though you consider more of a binary pattern of gene expression).

Answer 1

[Guangchuang Yu]

using this package: http://www.bioconductor.org/packages/devel/bioc/html/clusterProfiler.html

Answer 2

[Istvan Albert]

The way you derived a list is not relevant, expression levels are just one way to create one list versus another. You could group genes by name or page numbers by which they were published first.

You can always do an analysis - and if done properly the results would tell you about enrichment of one list versus the other.

Of course even if you were to find enrichment that does not necessarily imply a biological mechanism at play, it may just be that the the selection of the genes were biased in one way or another. To take our example to extreme genes that are blood related may be published in thicker journals than say those relating to development therefore even selecting genes by page number could show enrichment. The important part is that the enrichment that you would find in our example correlates with the way papers are published and not the actual biological function of the gene.

Answer 3

[Russh]

Are you studying a single cancer cell line versus a single 'normal'; or do you have multiple lists?

There's no reason why you can't do the enrichment analysis, on a single list (you will need to define a background list against which to compare). However, I suspect the enrhiments will be skewed by proteins at the 'difficult to detect' limit. May be interesting to compare to mRNA expression data for the same cell lines (if poss) to see whether you isolation method misses specific classes of proteins that have detectable mRNA expression.