Microarray Probe Expression Questions

Question

[yatung]

Recently I am doing the analysis of comparison between microarray data and RNA-seq data.

There is a problem confused me a lot...

The array data I used is from Affymetrix Human Genome U133 Plus 2.0 array. The gene names I used are from refseq.

Affy do provide a mapping table from probe to refgene ID. But I find out that the mapping table is not one to one...

That is to say, one probe might correspond to several refgene IDs.. And one refgene ID also will corresponds to several probes.

My problem is that which expression value of the probe should I pick for the specific refgene ID?

Or, how to combine different probe expression data to represents the specific refgene ID expressions?

Answer 1

[Darren J. Fitzpatrick]

How to pick is really rather subjective. If, indeed you need to pick. Can you not incorporate your many-to-one reportor to gene mappings in your comparative analysis of RNA-seq and microarrays. If not, you could apply arbitrary criteria to remove the redundancy.

1. You could take the median/mean of the multiple expression probes to represent that gene.
2. Take the probe that represents the longest/shortest splice variant.
3. Select the probe with the least number of outliers as measured by standard deviations.
4. If you have the probe sequences, select the probe least likely to cross-hybridise, i.e., the probe with the least similarity to other areas of the genome based on a BLAT search using UCSC genome browser.

These are just a few suggestions. It may be worth checking the correlations amongst multiple probes that map to a single gene. If they are all highly correlated (which may not be the case - damn noise!), I am guessing that it then won't really matter which you choose.

Good Luck!

[yatung]

Thans for your answers!! That's really helpful! The expressions of many probes to mapped to the same gene is really annoying... I'll try all of them and see which is better~ Thanks a lot!