It’s clear from recent progress that DNA sequencing will soon become commonplace in medical care – it already is widely used in biomedical research. DNA sequencing costs have dropped and will soon be at $1,000 or less for an entire human genome. The cost of interpreting that data is being debated but it will no doubt drop as well. There are some important things that need to happen before sequencing for medical use can become commonplace. I’ll discuss a couple of the most critical ones below.
The first is the issue of FDA and other regulatory approval. Right now, labs that do sequencing for clinical purposes at the very least have to be CLIA (Clinical Laboratory Improvement Amendments) certified. 23andMe’s genotyping lab (which is actually a LabCorp facility) is CLIA certified, and some sequencing service companies are achieving this certification as well. My understanding is that CLIA is about ensuring good laboratory practices for things like sample tracking and quality control. Even with CLIA certification, however, errors can happen.
But CLIA doesn’t really address how the laboratory device actually operates, or how the lab results are interpreted. FDA approval of a sequencer is still something that could happen, and whichever company achieves that first will have a big advantage. Illumina is certainly thinking about it, and Life Technologies announced that they’d be pursuing 510(k) clearance for their Ion Torrent PGM sequencer in 2012. Industry watchers will no doubt be handicapping that race.
But probably the most interesting part of clinical DNA sequencing is the problem of interpretation. There have been several recent research papers that have demonstrated how this can be done. But these academic tools need to be transformed into production systems that can be run reliably in high throughput environments, and the issue of differing interpretations from different tools will need to be addressed. And interpretation happens at several levels, each of which must be well understood and standardized before clinical use of sequencing.
First off, the calling variants by comparing a genome sequence with a reference is still not completely accurate. I’ve heard anecdotally that different algorithms can sometimes produce quite different results. And then there’s the issue of how you interpret the variants. This, I suspect, will be the most troublesome task of all. The data about what genetic factors are associated with which diseases and conditions is largely in scientific papers. There are quite a few organizations, both academic and commercial, that are building structured databases of these variants. But how will consumers of these results, both doctors and patients, understand the differing results they might get from different interpretation tools?
I’m sure there are other issues, and there’s plenty of room for debate about how and when DNA sequencing will touch the clinical care of a large part of the population. I’m looking forward to exciting developments in the next year.
-Will FitzHugh, 5AM Solutions