23andMe already testing for rare Parkinson's mutations?

This casual aside on a recent post on personal genomics company 23andMe's corporate blog caught my eye:

Mutations in several other genes have also been associated with
Parkinson's disease, but these are extremely rare. Many have been found
only in one or two families. While these mutations are so rare that
they are not covered by 23andMe (to date we have found no customers
with any of them), studying them could help scientists better
understand the mechanisms of Parkinson's generally... [my emphasis]

In other words, the company already has probes on its custom chip targeting these variants, but it isn't yet reporting results back to customers.

Why isn't it reporting back? If you'd asked me a couple of months ago, I'd say the motivation was probably to avoid the regulatory hassles associated with testing overtly clinical markers - but the company's willingness to provide results for large-effect variants associated with breast cancer pretty much rules that out.

Instead, the most likely reason to hold back on giving results back to consumers is (perfectly reasonable) caution about the reliability of the test. Screening for extremely rare variants is tricky for two reasons: firstly, since there are very few individuals around who carry the mutation, obtaining positive controls is difficult; and secondly, screening accuracy needs to be extremely high to keep down the rate of false positives.

To illustrate that last point, let's say there was a genetic variant with a population frequency of just 0.1% (1 in every 1000 people carry it)*. Now, let's say you have a test with false positive and false negative rates of just 1 in every 1000 tests, and you run that test on one million people. Of the 1000 carriers in the population, the test will only miss one; but it will also give a positive result for 999 people who are non-carriers. In other words, even for this extremely accurate hypothetical test, only 50% of the people who test positive are actually carriers.

This means that testing for rare variants requires exceptionally high standards of accuracy, probably higher than could reasonably be expected from chip-based assays. Given the risks of reporting potentially unreliable results back to customers for serious risk variants it makes good sense for 23andMe to hold off until it has developed extra assays for quality control; and it's unlikely to do this until it has seen at least a few customers who actually do test positive for the variant in question.

As for obtaining samples from real carriers to enable the development of validation assays: what better way to do that than to recruit 10,000 customers suffering from Parkinson's? Targeted recruitment of customers with other diseases will no doubt follow.

It is now abundantly clear that 23andMe is intent on moving into the overtly clinical domain; Navigenics' purchase of its Affymetrix testing lab and deCODE's move into disease-specific genetic tests are other signs that this is a shift that will involve the entire personal genomics industry.

Personal genomics is getting serious.

Update 24/03/09: 23andMe blogger ErinC has edited the post to read:

Mutations in several other genes have also been associated with
Parkinson's disease, but these are extremely rare. Many have been found
only in one or two families. 23andMe is not able to detect most of
these rare mutations.

And was kind enough to send me an email confirming that "[23andMe] won't be reporting any data to customers (about anything) unless we have the necessary positive controls for validation."

Subscribe to Genetic Future.

* I know, I know, if the allele frequency is 0.1% it will be carried by one in every 500 people. If it bothers you just pretend I said 0.05%.