On the road to personalized medicine

(A quick note on the term "biomarker": it sounds fancy, but it's just a term for something that you can measure to determine risk of disease. Doctors can look at cholesterol levels, lipids, etc. to predict health, but they hope to find better indicators of disease and develop more sophisticated tools to make precise measurements. Proteins and metabolites in the body change depending on a patient's disease state - finding one of these or a bunch that can accurately predict a patient's state would be a great tool for determining treatment.)

There are plenty of examples of disease states where earlier biomarkers would be incredibly useful for diagnosing early and preventing damage, but the kidneys are a particularly nice example because it should be fairly easy to test for a marker -- just take a urine sample or use a dipstick. Testing for KIM1 along with other "qualified" biomarkers might yield a robust diagnostic tool.

I've spoken to researchers at the Broad who have also been searching for biomarkers to help diagnose a condition early, but instead of looking for a protein or metabolite associated with kidney damage, they're looking for a marker of myocardial infarction (heart attack). If physicians are confident someone has just had a heart attack, they can take swift action, but it's not always easy to know. One of the difficulties that researchers at Broad wrestled with was determining a baseline level for these metabolites and proteins - you need to know what normal is before you can figure out what's off the charts. For heart attacks, there's a clever way to do this. There's a heart condition that is treated through a technique called septal ablation - as part of this operation, surgeons must induce a planned heart attack. The clinical researchers could monitor levels of proteins and metabolites in the blood before and after the planned heart attack and determine the exact timing when these markers appear as well as the change in the levels of biomarkers.

If the researcher had been here, I would've wanted to ask how he goes about establishing "baseline" levels for kidney biomarkers - also, does this level differ between individuals?

The session's organizer -Donna Mendrick -- was able to clarify the difference between two important terms that kept cropping up: "validation" and "qualification." When researchers talk about validating a biomarker assay, the term means making sure that the means used to measure a biomarker today will yield the same results tomorrow, next year, etc. If the assay itself is flawed or inconsistent, the data are worthless. A biomarker is "qualified" through clinical trials to make sure that the marker is telling you biologically what you expect it to be telling you.

Ivan Rusyn of UNC Chapel Hill spoke next about modeling toxicology. The first talk focused on detecting biomarkers for everyone - the second talk focused on understanding the differences between individuals and why a drug might work well for one person, but have horrible consequences for another.

Dr. Rusyn pointed out that we shouldn't just be interested in how the drug affects the average individual; we also need to know how a drug will impact the most susceptible people. This is of interest both to drug companies looking for the safest drugs and government organizations assessing toxins that we are all exposed to. Genotyping individuals is an expensive proposition so Dr. Rusyn suggests improving animal models to accurately reflect the genetic diversity in the human population.

Dr. Rusyn used the example of Tylenol, which can cause a gradient of toxic effects on the liver (i.e., toxic in some, non-toxic in most individuals). With too small a sample to perform genome-wide association studies and too large an amount of genes involved to pick candidate genes, Dr. Rusyn suggests that using mouse models could be an effective path forward. In the Tylenol example, the researchers were able to use mouse models to find candidate genes and conclude that response to inflammation determines toxicity response. Of course, this kind of model does not help determine epigenetic differences.

During Dr. Rusyn's talk, I was thinking about Warfarin - I guess my brain was just a few minutes ahead because Dr. Dennis Vargo, the next speaker discussed the drug. Warfarin (aka coumadin) is a blood thinner, a version of coumarin, all of which are anti-coagulents. Warfarin is given to patients at high risk of blood clots. Dr. Vargo noted that coumarin was actually discovered 125 years ago in Canada - cattle eating rotten clover began developing hemolysis (rupture of red blood cells?). People knew the drug had anti-coagulent properties and found a great use for it: rat poison. Warfarin today is a powerful but potentially dangerous drug for humans - the right amount can be lifesaving, the wrong amount can be deadly. Physicians must carefully choose the right dosage for a patient, but the right dosage varies widely depending upon a variety of factors including age, BMI, what food the patient eats, and genetics. (The name "Warfarin" comes from the Wisconsin Alumni Research Foundation - the "-arin" come from coumarin.)

Researchers discovered that CYP2C9 controls how warfarin is metabolized - this determines if a smaller amount of the drug is necessary to prevent excessive bleeding. Dr. Vargo posed the question, is the clinical benefit of routine genotyping enough to outweigh the cost? A member of the audience suggested that at the rate things are going, it's only a matter of time before genotyping is so cheap, that it will be cost effective to screen for those at high risk from prescriptions. Another audience member asked if similar studies were being done on aspirin - let's hope so.

The talk ended with a note about complex diseases and simple tests - a "Renastick" test might be enough to determine kidney toxicity, but predicting an adverse reaction to Tylenol might prove to be far more complicated. Overall, a great talk with really interesting examples, engaging speakers, and insightful questions from the audience.