Tuesday, December 26, 2006

BC on Autism 24: Microarray, Schmicroarray


Click to Play

A quick look at a paper which allows me to launch into my favorite criticisms of microarray data. In short, most people doing Affy arrays don't have any idea what they're doing. Good for Affy. Bad for science. Bad for the public. These guys even try to pull in autism into their analysis. Blech.

Running time ~7min

3 Comments:

At 12/27/2006 1:51 AM, Blogger Camille said...

So as far as this data went they didn't find any difference between the gene expression in Tuberous Sclerosis (type 2) folks with autism and without? Would you expect that that would mean something interesting IF they had had a larger sample size and separated the groups by age and sex and cell type? I mean it seems like it would be interesting... but it looks like they were a long way off from producing anything too meaningful doing what they were doing.

 
At 12/27/2006 9:17 AM, Blogger notmercury said...

"Good for Affy. Bad for science"

Well AFFX stock has been hammered lately, probably due to the early frenzy over their gene chip technology and the ultimate recognition of the limitations to the technology and saturation of the market.

Isn't the TGEN autism study in AZ using Affy products?

Have you read anything about this "genomic profiling" thing that the the DAN! folks are using?

 
At 12/27/2006 3:56 PM, Blogger Bartholomew Cubbins said...

Hi Camille,

The idea of this experiment is interesting. The authors ran into a practicality limit: simply put, a larger sample size gives better data but costs more. So while the chromosome 21 skewness for DS is interesting, one should not expect a whole blood assessment to work for any individual neurologic condition or disorder. The other thing that disappointed me was that these researchers should know that red blood cells have very little RNA and because of that, any increased representation of transcripts involved in inflammation or the immune system should be expected given the abundance of, the amount of RNA in, and biological roles of the white blood cells.

What speaks volumes is that two years later, no validation has been published IIRC.

The reason I did this paper was to highlight the frenzy, as NM aptly puts it, of microarray studies. These experiments are easy enough to accomplish that anyone with a pulse can knock one out. But it's harder to sit down and plan an experiment such that sense can be made of the results. Even harder to do competently is the actual statistical analysis of the microarray data.

NM - I agree heartily. Affy has a great product and I know that they've sunk a major amount of money into the new splicing arrays, so maybe it'll bounce back up as more people look at alternative splicing.

Re TGEN: don't know, but I'll be more than happy to rip them a new one if they put out lame data like this paper did.

Re DAN!: don't know. But from what I've seen, there isn't a DAN! doc out there competent enough to perform microarray analysis her/himself. They have enough money to buy the equipment for sure, but there's a skill deficit. What they do is send tissue (blood) out to a core facility that would run the sample prep and analysis for them. By analysis, I mean a very rudimentary handling of the data, which is troubling in and of itself. If they rely on a lab that does a Southern blot, well that's a yes/no answer if done right (think fragile X test). What I've read from DAN!ites is more of a SNP analysis which if they're using the 100K-type array from Affy, the mean distance between markers is 26 kb. I'm sorry, but that's some serious distance. I also have a problem with some traffic light (Red, Yellow, Green) lab report that also notes "mutations". I mean, how many mutations are conservative and undocumented in the literature? To sit back and dose a kid with SAM or MeB12 or something because someone tells you there's a mutation in gene X is pure guesswork. What's the mutation? What's it doing to the protein? Was there a Western blot done to verify a down regulation of protein X? I mean, if you're really going to be a metabolomics-genetics-expression doctor, you better close the loop. It's pathetic to go halfway and stop because you can convince a parent to open his/her own wallet on that data alone. Hey, but that's state-of-the-art DAN!. Be proud.

 

Links to this post:

Create a Link

<< Home