Login to Your Account

Human genetics no cure-all, but good medicine for drug development

By Anette Breindl
Senior Science Editor

Of the many types of evidence that a protein might make a good drug target, human genetic studies linking its gene to the disease being targeted is considered one of the better ones.

There are certainly examples that bear out this hunch – statins, for example, target the enzyme HMG-CoA reductase, and genetic variants in the gene for HMG-CoA reductase affect cholesterol levels.

Still, for now those examples are essentially anecdotes. And there are certainly counterexamples, such as Alzheimer's disease, where targeting the formation of amyloid beta has clear genetic support, but more than a dozen late-stage clinical trials whose rationale was based partly on that genetic support has failed. (See BioWorld Today, July 12, 2012.)

Now, researchers at Glaxosmithkline plc have taken a systematic approach to evaluate just how encouraging such genetic support should be. In experiments published in the June 29, 2015, online issue of Nature Genetics, the team compared drugs whose efficacy is suggested by human genetics research with those without such evidence. The researchers found that drugs with such evidence were about twice as likely to succeed in clinical trials as those without.

Beyond the exact number, the paper marks the first systematic assessment of the predictive value of studies showing the association of specific human genes with a disease for drug discovery aimed at those genes.

"We ask a very relevant question, which is 'how could human genetics research affect the success rate of drug discovery?'" corresponding author Matthew Nelson told BioWorld Today. His team's approach "lets us estimate what that impact would be."

The findings do not mean that focusing more on drugs with human genetic support is a cure-all for what ails drug discovery.

"We aren't going to improve our attrition rate to zero," Nelson said. Given the low overall success rate of drugs, making the whole enterprise twice as successful could also be described, albeit less impressively, as bringing the success rate up by 10 percent.

Still, given the high cost of drug discovery, such an improvement would make a real difference in the average cost of drug discovery. Since successful drugs have to pay both for their own development, and that of those drugs that failed en route to approval, another way of looking at the numbers is that doubling success rates would mean that each approved drug would have to be only half as successful to justify the entire enterprise.

In their work, Nelson and his colleagues conducted a retrospective analysis of approved and experimental drugs in different indications by drawing on several databases, including commercial databases that have drug and genetic information, and public databases of rare and common disease-linked genetic variants.

They found that across 16 different categories, drugs that were developed against targets that were linked to an indication by human genetic evidence were twice as likely to succeed as those that did not have such supporting genetic evidence.

The percentage of pipeline targets that had supporting evidence in the form of genetic associations for similar traits varied widely from more than a third in musculoskeletal and metabolic disorders to zero for the eyes, skin and connective tissue, and digestive system drugs.

Oncology drugs overall also had close to zero genetic support for their targets, a finding that might seem surprising at first. Nelson explained that his team's study did not include any data on somatic mutations, which is the cause of the vast majority of cancers. Overall, he said, the data supported the conclusion that "other things being equal, we should focus on drugs with human genetic support" to maximize the chances of drug development.

It is also true, however, that most successful drugs do not have genetic evidence in their support, for two reasons. For one thing, given the overall number of populations, and the overall number of diseases, that could be studied, "we really are only just scratching the surface of how genetics affect disease. . . . We are nowhere near the end of the genetic association era."

For another, an association between genetic variants and disease risk can only be found if such variants exist in the first place. A gene might be related to disease without there being a mutation that flags it as such, or, as Nelson put it, "not every gene is going to have a natural experiment." //