Wouldn’t it be wonderful if we could go to the doctor, have a few tests, and then be told which of the many diabetes treatments would work best for our particular physiology?

That’s not likely to happen in the next week or month or year or so, but it will probably happen sometime in this century.

An interesting poster presentation  at the recent annual meeting of the American Diabetes Association described how one company, Entalos, was able to predict that a certain drug wouldn’t work well in humans, although it would work in mice. (You’ll have to register to see the poster.)

The drug they simulated blocks an enzyme called glycogen phosphorylaes (GP), which breaks down glycogen in the liver and produces glucose, which can then be released and makes blood glucose (BG) go up. But which drug was studied is not as important here as the fact that their approach seemed to work.

First they created a group of virtual patients by using data on human physiology from thousands of published papers. Their virtual patients were not identical but, like real patients, varied a bit in their physiology. Hence when they simulated an oral glucose tolerance test (OGTT) on the virtual patients, they got a range of responses. Then they simulated what would happen if they gave them metformin. Then the study drug, the GP inhibitor.

In their model, the metformin had a significant effect at a range of initial hemoglobin A1c levels. The higher the initial A1c, the greater the effect of metformin. But the GP inhibitor had very little effect in the virtual humans (see figure 5 on the poster).

Then the researchers  created a bunch of virtual rodents (much less smelly than the real ones) and did the same manipulations. In this case, the GP inhibitor had a significant effect. This had already been shown in live rodents, and that validated the approach.

Using their data, they then hypothesized why the GP inhibitor worked well in rodents but not in humans.

This type of research could have several benefits. First, we’ve all heard of promising drugs that cure diabetes in rodents but do almost nothing in humans. Instead of spending millions of dollars on human trials, drug companies could simulate the effects of drugs that helped rodents to see if it was worthwhile to do the human trials and then spend their money on the drugs that promised the best results. Those that showed promise in the virtual patients could be tested in real humans.

They could also choose different populations of virtual patients to see if the drug was more likely to work in certain types of patients than in others. For example, if there was no good drug for treating patients with diabetes and some coexisting disease, they could run a simulation on that population of virtual patients.

These approaches would reduce the cost of development of new drugs and, one hopes, the costs of the drugs, although it might just result in higher salaries for the CEOs of the drug companies.