The fairytale of common references in Microarrays within SMD...
Markus Krupp
Tuesday, 22 April 2008 13:25 UTC
...is there a happy end for me?!
Hi everyone!
I’m Markus, I have a problem and hopefully nature network can help me.
...YOU can help me :)
The problem is concerning my Microarray dataset of human tumor tissues downloaded from the Stanford Microarray Database.
One step in my research is to build up a strong database of over-/underexpressed genes within human tumors, therefore I contacted SMD-FTP and downloaded involved MA data. The experimental design off all downloaded Microarrays is based on competitive hybridization paradigms using a pool, in other words …a red green array.
However “breast carcinoma, ductal, grade 2” for example has 42 entries in my dataset and those 42 entries reference to 7 different reference pools used in the MA experiment. My thought is either to remove those data or just to remove those data whose references are not prevailing equal. Case 2 would be much better for my analysis.
Look up SMD and vast amount of literature did not reveal any information about organization and participation on gensets within the “common references”. But this information is essential if I want compare MA experiment 2154 with MA experiment 3165 for differential genexpression for example (link 2154 link 3165 ).
Below is the list with used “common references” within my dataset collected from SMD where I need further information.
- CR1 Reference
- CR2 Reference
- CR3 Reference
- CRB1 Reference
- CRD Reference
- CRE Reference
- CRF Reference
- CRG Reference
- CRG Reference + Placenta
- CRH Reference
- CRS Reference
- Normal female genomic DNA
- Stratagen Reference cell lines
- Stratagene Reference + skin samples
- Stratagene Universal Human Reference
- Stratagene Universal Reference
- Universal Human Reference
Thanks for any help!
BW Markus
Updated 22 April 2008 13:25 UTC
-
Replies
-
The use of different common references does pose a big problem when comparing one dataset to another especially when you want to compare more than just two datasets. I’ll have to spend a little time looking into exactly what you are talking about here. Can you try to explain the problem a little more? Is it just a different references problem? I worked with the breast cancer data from the SMD several years ago (I did my PhD work in the lab of Chuck Perou). I think that Stratagen Reference cell lines, Stratagene Universal Human Reference, Stratagene Universal Reference, and Universal Human Reference may all be the same thing. I’ll look into this and try to hopefully answer it in the next week or so.
Also, not that its important for your question, but it is my understanding that this isnt truely competitive hybridization. There is supposedly more than enough probe present in the spots on the array for both the reference and experimental sample to bind non-competitively.
-Jason
-
Hi Jason!
This is my first time in analyzing Microarrays and therefore I appreciate your help very much.
I think that the Stratagene references used in SMD are the same thing (namely the Stratagene Universal Human Reference), but I can not prove it. On the ohter hand I find no information on the other references, whether in the literature or the Internet. I’m afraid that CR1, CR2, CR3, etc. may be “selfmade” and therefore not compatible to others.
To get further information I have sent an email to SMD and today I will contact the workgroups doing those arrays.But here we go as computational biologists and say there is more need for a Microarray ontology and standard. MIAMI, MAGE, ... are great and well implemented but we need more :)
Maybe your right with your understanding of competitive / non-competitive binding to the arrays, at the moment I’m more informatics than biologist but I keep going. And such discussions maybe lead me to an balanced milieu :)
Thanks again,
Markus -
Just for information:
Stratagene supports just one reference set. Therefore it is completely valid to regard the Stratagene references as the same thing.
However is still do not know what’s behind CR* references. I presume that those references were “self made” and on that score they will not implemented into my dataset.
-