I am looking for genes that have altered expression levels. When I use the GEO database I find multiple results for the same gene within one dataset (for example MCM4 in dataset "GSE9750"). In this example MCM4-results appear 4 times: 2x MCM4 is significantly elevated and 2x MCM4 shows no significant elevation. I'm wondering why one dataset contains multiple results for the same gene that also appear to contradict each other.
I appreciate your help!
As expected, this is microarray data. Usually one probe is used to detect one gene (transcript) in an array but it's also common that some genes have multiple probes designed to hybridize to different regions of the transcript, perhaps for the purpose of variant/isoform studies. If you are interested in the overall expression of the gene, I would usually take an average of all probes for that gene. Otherwise you could look up the information of the probes (their locations, etc.) from the array company. In your case, http://www.affymetrix.com/estore/catalog/131537/AFFY/Human+Genome+U133A+2.0+Array#1_1.
As expected, this is microarray data. Usually one probe is used to detect one gene (transcript) in an array but it's also common that some genes have multiple probes designed to hybridize to different regions of the transcript, perhaps for the purpose of variant/isoform studies. If you are interested in the overall expression of the gene, I would usually take an average of all probes for that gene. Otherwise you could look up the information of the probes (their locations, etc.) from the array company. In your case, http://www.affymetrix.com/estore/catalog/131537/AFFY/Human+Genome+U133A+2.0+Array#1_1.
As Menglan Xiang said, in an array you can have multiple probes which correspond to the same gene. However, I don't agree for averaging the expression values of all probes for that gene and here an example: We suppose that for one gene you have four probesets with expression values -2, -1, 2 and 1. The average of all probes will be 0 which corresponds to a non significant expression variation of that gene between the two conditions.
Thank you for your answer Kais. In your example, how would you select which probe results to "trust"?
Not easy question :) What I am doing in that case is to: 1) select the probeset with the highest absolute expression value. In my previous example it will be 2 with two probesets (-2 and 2). 2) then to see in the literature if this gene is found up or down regulated under the same condition. You can also do a functional analysis, pathways enrichment to identify the most significant pathway in which your gene is involved and see the role of this gene in that pathway (if it is more likely to be up or down regulated). I hope that my answer is helpful for you.
there are multiple strategies how to deal with more than one probe results.
-the higher expression levels
- average median
neither one is perfect. However, if it boils down to just one or two genes deeper investigation is required to check which exons are targeted by what probe. it is quite frequent that differences are explained by differential expression of alternative splice variants of the same gene
Hi Sebastian,
To solve this problem you should check array probe id in particular it's suffix. Affymetrix come with several different types of them for example 123456_x_at which describe probe to target affinity. Based on them you can build a priority list of probes from most specific to least specific. Of course best ones are those most specific but unfortunately it is often impossible to develop unique target for every transcript so in order to detect anything developers put some less specific probes. So if you use less specific probes and average your results you introduce more false positive results and in case of that you should verify your result by other methods (Nothern Blot, QPCR). For more info about suffixes check link below. In order to calculate final expression you could discard less reliable probes or introduce weighted average to distinct more reliable probes.
http://www.affymetrix.com/support/help/faqs/mouse_430/faq_8.jsp
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology