Microarray Expression For Genes With Multiple Probes

Question

[tintin123]

Hi all,

I am working on microarray expression. I am new to this, so pardon me if I provide incomplete details.

I have downloaded the raw data from GEO (gilent-014850 Whole Human Genome Microarray 4x44K G4112F) and we want to look at expression levels of genes with different datasets added on top (expression levels of genes with and without epigenetic marks, transcription factors).

When I was extracting the information, I realised that there are multiple probes for certain genes. At first, I took the highest expression values for each gene and then later compared it with the average and the difference was notable.

I was wondering if there is a method to identify which probes are present in all isoforms and which are present in a few.

Thanks for the help.

Answer 1

[Jelena Aleksic]

If you get the probe sequences, you should be able to map it back to the genome (a useful exercise anyway - sometimes the genome sequence has changed and the probes e.g. no longer map uniquely), and then you can use the latest gene annotations to figure out which specific exon they hit. From there, if you sort by transcript ID, you should be able to get an idea. There also seem to be some alternative splicing databases in existence, but I've never used them, so can't tell you anything about them: http://www.eurasnet.info/tools/asdatabases

[tintin123]

Thanks for the reply. This is a good idea. I will give it a go.

Answer 2

[Charles Warden]

Yes, I agree with mapping the probe sequences on your own. However, I typically wouldn't do this prior to analysis - in practice, there are a lot of probes to check.

I would typically conduct differential expression with all the probes, and allow a single probe to be sufficient for differenital expression (at which point, I could take some time to understand any significant differences between multiple probes that map to the same gene - for specific results that look interesting).

Your strategy should depend upon your method of integration. If you are looking for overlapping gene lists, I would recommend the strategy listed above. However, if you need to work with the absolute expression values in the integration process and you need a single expression value per gene, I would typically just average the expression among probes that map to the same gene.

[vipin.ranga]

Thanks for the information. Could you please refer me an article which provides information about averaging of expression values among probes that map to the same gene? I will be very thankful to you. Regards

[Charles Warden]

I don't recall any publications off the top of my head - as I mentioned, I typically consider one probe to be sufficient to define differential expression. I'm guessing that a lot of genes have 2 probes, in which cause you'll pretty much have to use the mean as a summarized expression value from both probes.

[vipin.ranga]

Yes I am agree with you, I am also using the same concept but the thing is that If I am using this particular approach then I must have to mention the reason to answer the question of averaging of expression values. I am afraid if someone ask me about it then I must have evidence to make them sure.