The cost of sequencing a human genome has dropped to just below $5,000. Well, sort of. The Mountain View, Calif.-based company Complete Genomics published a paper today describing its efforts to sequence three genomes at a materials cost of $4,400. Included in the trio is the genome of George Church, personal genomics evangelist and an adviser to the company, who has already posted the analysis of his data here.
When Complete launched last October, it promised to sell 1,000 $5,000 genomes by the end of this year. The company’s business model is new in the sequencing world: it performs sequencing and analysis in-house, unlike other companies, who make money by selling sequencing instruments and reagents to customers. And it only sequences human genomes.
The big question when Complete launched was: could it deliver? This paper answers that question: Yes, the company can sequence human genomes. And yes, they are cheaper that those recently sequenced by other companies. But no, they’re not selling them at $5,000 per genome just yet, and it’s a bit difficult to interpret the quality of the genomes described in this paper.
Complete takes a bit of a brute force approach to genomics. Like many other companies its technology assembles small pieces of DNA into whole genomes. By sequencing each genome enough to read all of its bases an average of between 45 and 87 times, Complete hopes to eliminate most of the errors that result from trying to put together giant jigsaw puzzles consisting of tiny fragments of amplified DNA. (By comparison, other individual genomes have been sequenced at anywhere from 7 to 28 times average coverage.)
The company said that it sequenced one of the three genomes in this paper over twelve days using three machines, then assembled the resulting data into a genome over two days on a 600-core computer cluster.
So, are Complete’s genomes, well, complete? The company reports that it can correctly determine 94 to 98 percent of the genetic variants that it tested per genome. Is that good enough? In September, at the Personal Genomes meeting in Cold Spring Harbor, NY, one of Complete’s first customers estimated its error rate at 10 to 20 errors per megabase. The current paper reports a “false discovery” rate of one bad call per 100 kilobases, or 60,000 bad calls per genome. But the company used a method that confused most genome scientists queried by Nature to calculate this number, so it’s difficult to know how that compares to other technologies. The new paper also doesn’t address issues like, how well does the technology discover the large structural variations, such as inversions, deletions, repeats, and large rearrangements of DNA, that are important to human health and disease?
The real question is: is this good enough to pinpoint one single mutation that might underlie a patient’s disease? At the September meeting, Complete’s customer said that it had detected three potential candidate mutations in two children with a rare disease called Miller Syndrome, and that the causative mutation was on that list. So at least, in this case, it seems the technology worked.
The question now is whether this data will convince scientists, biotech companies and pharma giants to outsource genomics services to Complete, thus enabling Complete to further cut its costs through economies of scale.
One Complete customer, Chad Nusbaum of the Broad Institute, says the data in the new paper are roughly comparable to the data from one Complete genome he has been able to analyze so far, and that, “in principle,” the Broad would order more, depending on the price and turnaround time.
The Institute for Systems Biology in Seattle, run by Lee Hood, a Complete adviser, was the customer that ordered the Miller Syndrome study, and has ordered a batch of 100 more genome sequences to investigate genetic traits that modify Huntington’s Disease, for which the causative genetic glitch is already known.
At least one pharma company, Pfizer, has been publicly identified as a Complete customer.
And Church has enlisted Complete to sequence the initial 10 participants of his Personal Genome Project . (Asked what he had learned about his own Complete genome, Church told Nature, with characteristic wry humor, “PGP1 is a very boring person/genome.”)
Complete’s $4,400 price tag doesn’t include costs for the company’s infrastructure, such as its Silicon Valley data farm and the army of analysts and technicians required to make sense of the data; the company lists more than 60 employees in this paper’s author list. The company is actually selling genomes at $20,000 apiece in minimum orders of five; costs go down as the order size increases. That puts it slightly behind the schedule it set at its launch; the $5,000 genomes won’t be available until next year.
The bigger point here, however, is that sequencing costs are finally dropping low enough so that the technology can be used to discover the genetic roots of patients’ diseases. Two other scientists at the Personal Genomes meeting in September also reported that they had used different next-generation sequencing to pinpoint disease-causing mutations in patients. (One of those cases has been already published here). Meanwhile, dozens of new sequencing companies are working away in garages and office parks on ever-better technologies that will further cut costs and timelines. Personal genomics has been a long time coming, but if these companies can deliver, we shouldn’t have to wait too much longer.