Heyo -- thanks for the heads-up on Twitter. I'm the sysadmin at a small university department, and I work with scientsts studying gene expression. They're good and patient people, but sometimes I feel a bit like I'm questioning the foundations of their work...which feels either rude or ignorant.
First off, I'd always been under the impression that DNA was only/mainly used during reproduction -- a cell divides under DNA direction, some bit of the cell is the machinery that makes whatever protein is needed during its life, and DNA isn't involved much after that. However, I'm starting to understand (I think...) that I've got it all wrong. My understanding now that gene expression can basically turn on a dime, and that *this* is the usual way a cell makes a protein: something happens to a cell, it says "Whoah, I need protein X", and it starts transcribing the DNA so it can manufacture it (modulo things like gene regulation). This process can take very little time (hours or less). Have I got that right?
Second: one of the things they study is datasets of gene expression in post-mortem brains. (Well, technically I guess I've got that wrong, since genes aren't expressed post-mortem... :-) As I understand it, someone dies -- say, someone with schizophrenia -- their brains are donated to science, and at some point someone does microarray sequencing of blendered neurons. This is compared to brains of control subjects, gene X is found to be over/under-expressed in schizophrenic brains, and so gene X is involved somehow in schizophrenia. (This is a gross simplification, especially in the case of schizophrenia; my understanding is that these signatures cover many, many genes, they're subtle at best, and there's nothing like "a gene for schizophrenia".)
What I don't understand:
a) Since time passes between death and sequencing, how much fidelity does/can this have do what was going on at the point of death?
b) Even if it is a good indication of what was going on at death, how does that relate to a long-term illness like schizophrenia when (assuming I've got this bit right) gene expression can turn on and off in a very short time? I realize there are (ahem) ethical problems with doing brain biopsies on living subjects, and that post-mortem is the best that can be done -- but how good can it be?
Many, many thanks for your time. Any questions about system administration, let me know. :-)
Hope you don't mind me hijacking this thread, I think it's a great service Samantha is providing here. I just wanted to add a few comments as someone who has sadly seen alot of sloppy gene-chip experiments going on (but also some very nice ones).
It's really encouraging to hear that you are taking an active interest in what your scientific collaborators are trying to show. You'll be that much better equipped to help them prove what they want to show if you are roughly on the same page as them - something alot of scientists overlook when they delegate out the technical stuff they don't know how to do themselves. You might find that the group you are working with has some graduate students or maybe postdocs (i.e. probably whoever you have day-to-day contact with who actually does the experiments and hands you with the datasets) that would be much more available to answering your questions than the big bosses who have to consult a calendar to even see if they have time to meet with you.
As a biophysicist (but importantly not a neuroscientist), I can still say that I am not aware of any consensus on what causes schizophrenia other than it must have both a genetic and an environmental component (i.e. having relatives that had it greatly increases your risk, as does certain types of substance abuse). Therefore it is absolutely a central assumption of your collaborators' research that some key component is due to a long-term up or down-regulation of expressed mRNA's, that's certainly not an established fact anywhere in the literature although there is speculation and circumstantial evidence and it might be a favorite hypothesis in the field. I am 100% certain that a large portion of the grant proposal that funded this research was devoted to justifying this assumption and that it will be the first question out of the mouths of peer-reviewer of any papers that come out of it, so I can only assume that very persuasive arguments were made since microarrays are expensive.
I can give an example of an over-simplified hypothesis that, nonetheless, would be a home run to your collaborators if they could prove it. Certain drugs that interfere with neurotransmitter receptors (ketamines?) can induce schizophrenia-like symptoms, so maybe one component of schizophrenia is something that systematically lowers mRNA levels of the neutrotransmitters that interact with ketamines so there are less of them in the brain. It would have to be chronic and long-term to make a huge difference to cognition (surface receptor concentrations take along time to build up as they are expensive to make) so such a scenario would in fact show up in the mRNA levels. Or maybe something upstream that helps promote production of a neurotransmitter (like an activator) gets down regulated and has the same end effect. But I can also make up hypothesis that don't involve mRNA's at all - What if reduced neurotransmitter function is caused by some sort of protein misfolding due to a genetic mutation, so they end up being recycled instead of on the cell surface? Or what if it's an adverse reaction to something else in the environment that interferes with receptor function as opposed to merely diminishing their numbers? As Samantha pointed out, people measure mRNA levels mainly because the technology exists, whereas it simply doesn't for alot of other important processes.
Off the top of my head, there are super-lots of other assumptions that would have to be worked out as well, even assuming there is no significant degradation of the mRNA after death (which as Samantha points out, requires freezing to be absolutely sure, but AFAIK you don't freeze donated organs, since mammalian cells don't handle freezing and thawing very well). Will mRNA expression levels at death have more to do with the dying process than the underlying long-term neurological condition? Would it affect all neurons or just a certain type from a certain part of the brain? What is the false-positive rate on the gene chips themselves? (I've read alot of commercial ones have errors on them). These are all questions that a cell biologist/neuroscientist would be able to answer far better than I (i.e. the lab you are working with). But I would characterize your concerns about "questioning the foundations of their work" as totally legitimate questions since they are relying on you to help them sort the signal from the noise - a task that would be much easier if you know what assumptions are being tested by the controls and where to look first when you get a new dataset.
Good luck with your work!