Analysis Of Microarray Hits Using Bioinformatics Methods

Question

[Khader Shameer]

I am working on the analysis of a vascular disease based case-control gene expression study.

I am looking for some of the best-practice bioinformatics based meta-analysis methods to explain set of genes identified from a microarray experiment using bioinformatics methods. I have already tried enrichment analysis (GO terms and KEGG pathways) and literature mining which are routine in the literature. Is there any other bioinformatics method that I should try to understand more about the hits from microarray analysis ?

[Timtico]

Could you explain in a bit more detail what your research question is?

[Khader Shameer]

Timtico, details about research question added.

[Casey Bergman]

Khader, could you say a bit more about what text mining approaches you tried here?

[Khader Shameer]

@Casey: I haven't used any text-mining tools. We were looking at a specific set of vascular disease related terms. I integrated MesH, GeneRIF and Pubmed IDs linked to Entrez Gene.

Answer 1

[Larry_Parnell]

Khader, this is an excellent question and I have been giving this some thought on and off throughout the day. Researchers often full into ruts - doing analysis such as yours with GO and KEGG (or Reactome or similar pathways) just like everyone else who has presented at conference or in print, or writing grants in a similar style with similar trendy approaches for similar reasons.

So, is there something else you can do?

With the emerging data linking GWAS to eQTL, I would suggest looking at GWAS or similar genetic signals (could be from human GWAS or could be from mouse/rat data and their phenotype affecting, say blood pressure). This is not that different, admittedly, because it boils down to enrichment analysis, but from a different set of genes. Here is a great paper that shows that many cardiovascular disease GWAS signals are also eQTLs in pertinent tissues. Similarly, which rodent phenotypes would you find interesting? Go get those genes involved in that phenotype from the MGI or Rat genome databases and analyze for enrichment in your data. Imagining being able to say that of Y rodent genes whose knock-out/knock-down phenotypes show increased blood pressure, X of those human orthologs also show decreased expression in the data. That could be powerful if X/Y is high a statistically significant.

[Khader Shameer]

Thanks a lot Larry for sharing your thoughts. I have read through the Folkersen et. al before but never thought of an approach that you suggested.

Answer 2

[Andrew Su]

I'll only add that in addition to GO and KEGG, you might consult MSigDB for a broader collection of reference gene lists against which you can do enrichment analysis.

[Khader Shameer]

Thanks Andrew. I have heard of MSigDB, never tried it though - how different / overlapping is the MSigDB 'lists' in comparison with 'GO' or the 'pathway databases' ?

Answer 3

[User 59]

It almost pains me to admit advocating a non-free non-OSS solution, but I get a lot of mileage out of Ingenuity Pathway Analaysis for downstream processing of lists from array experiments/miRNA experiments. It's not fundamentally so dissimilar to GO/KEGG analysis, but it's backed with a lot of interaction data - capable of suggesting novel functional networks from your data, as well as providing a Cytoscape-esque interface for building networks around genes of interest, examining biological pathways in greater detail, and indeed just using it as an exploratory tool around a gene of interest. Obviously similar functions are provided by MetaCore and Selventa. Data export isn't so great from IPA, but there is an API of sorts (although I am informed Metacore is more amenable to programmatic interfacing)

[Khader Shameer]

Thanks Daniel - I will check those resources.

Answer 4

[Shigeta]

don't forget biocyc: http://biocyc.org/

[Khader Shameer]

Can you briefly describe the application of BioCyc from the context of microarray meta-analysis ?

[Khader Shameer]

Thanks Shigeta, can you briefly describe the application of BioCyc from the context of microarray meta-analysis ? – Khader Shameer♦ 0 secs ago

[Shigeta]

I haven't done it, but i'm starting to look over the annotations. i expect you will get good annotation data by taking the gene names or refseq accessions associated with the microarray probe sets and then attaching the biocyc data using their gene/accession assignments.

not sure which array you are using, but they all have some source of transcript/gene assignments for them.