AT A GLANCE
Neuroscience Investigator, New York State Psychiatric Institute, Columbia University Health Sciences Center.
PhD, Cornell Medical College department of pharmacology. Worked in Miklos Toth’s lab.
Postdoctoral fellowship with Rene Hen, Columbia University College of Physicians and Surgeons. Focused on animal models of depression.
Current studies focus on expression profiling of genes implicated in anxiety, depression, and suicide in mouse models and humans.
When did you start using the Affymetrix arrays?
Three years ago, during my postdoc focusing on animal models of depression, I was looking at candidate genes. Affymetrix had one mouse chip back then. I thought, ‘well, let’s try it.’ It was a total catastrophe because the machine kept breaking. It took me five months to get two miserable arrays scanned. And we had no idea what to do with the numbers.
Since you started, how have the arrays changed?
The quality has improved dramatically, as has Affymetrix’s approach to the clients: They were very secretive at the beginning, and now they are more open. They have a new website, NetAffx, that is fantastic: You can sit at your computer and just go from gene to gene, look through your list, and think. Things that would take hours a year or two ago are very much available.
I’ve talked to researchers about resolving issues of analyzing gene expression results between the U95 arrays and the new U133 arrays. How are you going to deal with that?
We are starting a new study with the U133s. When comparing between studies, the naïve view I have now is whatever individual gene is robust on one [the U95-chip] should still be there [on the U133]. What’s going to be a problem is large patterns, where there is clustering and large scale analysis. It will be most likely a different pattern. So we will have to be a little creative and prepare.
Another reason we are starting a new study is that [the Columbia Affymetrix core facility] is going to upgrade the photomultiplier tube (PMT) setting on their laser scanner. So even if I use the same chips, most likely I could not compare the data.
Why are they changing the scanner settings?
They realized that the intensity of the scanner had to be decreased by tenfold from the earlier chips to what they use now. For example, on some of the chips I have a big problem with saturation of signal with very, very highly expressed genes, which is to do again to the PMT of the scanner, so they have to decrease it. The problem is when they decrease it, everything scanned before and after will be quite different. If this happens in the middle of the study you are in trouble. We are trying to negotiate to get another scanner, so we can have one with the old settings and one with the new level. Because at a university, you may have 20, 30 studies at the same time. You cannot say “all right everybody, January 15 finish your studies. January 30, start another one.”
What sort of data analysis tool do you use?
I have a very small lab. I don’t have the resources to deal with the technical problems, because if I did that I wouldn’t have time to do my research. When the new Affymetrix Data Mining Tool 5.0 came out, I did an extensive comparison of versions 4.0, 5.0, and d-chip [the free software put out by Wing Wong at Harvard], and compared some of them by PCR. Version 4.0 needed a lot of tweaking, but 5.0 is very decent. D-chip may be a little more sensitive at finding low-expressed genes that were not found by the 5.0 and confirmed on PCR. But now the Affymetrix data mining system works pretty well so you can really use the data as it comes out, if you really understand the problems you are addressing, and you can plug it into other analytical software that is available. There is a lot that you needed to do in the other [versions] that you don''t need to do anymore.
With human RNA samples, sample integrity can be a big issue. How do you deal with that?
We get our human brain samples out of the Conte Neuroscience brain collection here at Columbia. And they take great care in the way they collect the brains. They’ve been frozen at minus-80 or so. In the study I have used brains that have been frozen for maybe ten, fifteen years. And the RNA is beautiful. Out of 40 samples that I studied, two or three were not good.
Is there any specific tool that you would like to see developed that would really help you in your work?
It would be nice to build a good relational database between genes with solid gene families and pathways, an easy and available tool that centralizes available data. It’s really a question of annotating the genome.