You can't always get what (samples) you want.

A key problem in omics research in medical research is getting the samples you need.

When I was an undergraduate, I had a fuzzy notion of a scheme for personalized medicine. Some analyzer would take a sample of what ailed you, look at it, and then generate a vial of customized antisense medicine that your doc would inject into you. I drove the pre-med in the lab nuts with my enthusiasm for it.

In graduate school, the analyzer became more clear: expression profiling. Look at the mRNA profile, figure out the disease, and voila, you are cured.

Fast forward to the latter part of my Millennium tenure. Rude surprise: you can rarely get the samples you want.

Most of my later work at Millennium was around cancer, originally because that is the research area I gravitated to & later because that was the one research area left (corporate evolution can be brutal!). Getting cancer samples turns out to be decidedly non-trivial.

If you are working in leukemia or related diseases (such as myeloproliferative syndromes), then things aren't bad. Your target tissue is floating around in the bloodstream & can be gotten with an ordinary blood draw. Patients in our society have been conditioned to expect lots of needle sticks, so this isn't hard.

For multiple myeloma and some lymphomas, you can go into the bone marrow. I'm a needlephobe, so the idea of a needle that crunches on the way in is decidedly unpleasant & sounds painful, and apparently is. Patients will do this infrequently, but not daily.

For a lot of solid tumors and other lymphomas, good luck -- particularly with recurrent disease. The tumors are hidden away (which is why they are often deadly) and quite small (if detected early). In many cases, getting a biopsy is surgical, painful, and perhaps significantly dangerous. You might get one sample; repeat visits are generally out of the question. Melanomas are one possible exception, but only for the primary lesion and not the metastases hiding everywhere.

This has significant implications. For a lot of studies, you would like to watch things over time. For example, what does the expression profile look like before and after drug treatment? How long does it take a pharmacodynamic protein marker to come up and what does its decay look like? Without multiple samples, these studies just can't happen.

Worse, what comes out may not be any good. Surgeons are in the business of saving lives, not going prospecting. Traditional practice is to cut first, then put away the samples after the patient is in recovery. But RNA & protein translational states are fragile, so if you don't pop the sample in liquid nitrogen immediately your sample may go downhill in a hurry. Multiple papers have reported finding expression signatures relating more to time-on-benchtop then any pathological state. It often takes dedicated personnel to perform this -- personnel the surgeons would rather not have 'in their way' (I've heard this first-hand from someone who used to be the sample grabber). A dirty not-so-secret in the business is that fresh frozen tissue just isn't practical for routine practice; you have to go with something else.

That is going to mean you go with several less palatable, but more available, options. One is to develop techniques to look at paraffin-embedded sections, which are the standard way of storing pathology samples. There are gazillions of such blocks sitting in hospitals, tempting the researchers. But, most of those sat on benchtops for uncontrolled time periods, so there may be some significant noise. Another is to try to fish the tiny number of tumor cells (or DNA) out of the bloodstream or perhaps an accessible fluid from the correct site (mucous from the lung; nipple aspirate for breast cancer). Or, you try to find markers in the blood or skin -- not where you are trying to treat, but easy to get to.

Whether these will work depends on what you are really looking for. For a predictive marker, it seems plausible that shed DNA or an old block might work. On the other hand, for a pharmacodynamic marker these are useless. A good PD marker allows you to measure whether your drug is hitting the target in vivo and at the correct site, and only by getting the real deal is that going to truly work. By necessity some studies use accessible non-tumor tissue, such as a skin punch or peripheral white blood cells, to at least see if the target is being hit somewhere. But that doesn't answer the question of whether the drug is getting to the tumor, a critical question. And many studies still use the traditional oncology PD marker of whether you are starting to destroy the patient's blood forming system.

At ASCO this summer, one speaker in a glioma section exhorted that a central repository for glioma samples must be imposed on the community, with a central authority determining who could do what experiments on which samples. That sort of extreme rationing shows how precious these samples are.

The scarcity of such samples also underlines why sensitive approaches, such as the nanowestern, are so critical. With small sample requirements, you might be able to go with fine needle biopsys rather than surgical biopsies, or be able to take lots of looks at the same sample (for different analytes).

Of course, things could be worse. What if you go to the trouble of getting a good sample, but then you look in the wrong place in that sample? Well, that's a post for another day.