Ultimately, however, somebody— perhaps a medical oncologist or geneticist—needs to sit down with the data and consider each novel variant to determine which, if any, might either cause a disease or represent an actionable target. This may involve researching the genes and digging into the literature, all of which takes time. And money.
Distinguished Scientist Gary Schroth, at Illumina, says bioinformatics costs largely accrue at this step. “That's where you have to consider the vastness of the human genome. Three billion letters is a lot to sift through.” Even after multiple rounds of filtration, the list of variants can number in the tens or hundreds. “You can't have a bioinformaticist or geneticist staring at that list for very long and keep the total cost under $1000,” he says.
To get a sense of the complexity of the problem, consider the Laboratory for Molecular Medicine (LMM) at the Partners Center for Personalized Genetic Medicine in Cambridge, MA. Director of Operations Lisa Mahanta has a single HiSeq 2000 in her lab, which she uses to run panels of 11–71 genes from as many as 80 patients at a time, searching for causes of deafness and cardiac problems, among other ailments. This is targeted sequencing and, according to Mahanta, “ridiculously small” in light of current genomics capabilities.
Still, it takes Mahanta's lab six weeks to return a 71-gene otolaryngology panel, and not because of backlog. The next-gen sequencing is done in about two weeks, and then it takes four to five days to pass the data through the lab's automated bioinformatics pipeline. (Data transfer alone—it's one terabyte of information—takes 24 hours.) The pipeline churns through the raw data, eliminating common variants known to be harmless or irrelevant. What's left, anywhere from 5 to 20 previously undocumented variants, now has to be scrutinized: first through independent validation by Sanger sequencing, and then with literature searches to determine potential variant impact.
It should come as no surprise, then, that the lab's recently launched genome sequencing and analysis service requires some time for turnaround. For $9000, the lab will outsource the genome sequencing to Illumina after which LMM personnel will interpret the resulting data, focusing specifically on “medical anomalies” with unclear genetic causes. All total, Mahanta anticipates taking four to six months to process each sample.
Of course, time and knowledge will speed the process: once a troublesome genomic variant has been identified and characterized, it never need be characterized again, notes Harvard geneticist George Church. In other words, once a mutation and its consequence (and potential treatment) have been defined, those details can be codified into an algorithm that can scan variant lists automatically.
Church's lab has developed a piece of software to go along with his Personal Genome Project for this very purpose. Called Get-Evidence, the software is a collaborative research tool in which researchers and physicians can enter information about novel genetic variants and their potential clinical impact. Once a variant has been added, “then no one has to do that again—at least until the next breakthrough in that gene,” Church notes. “The second person is basically free. And the third person and the fourth person; the downstream cases are free.”
Church notes that one participant in his PGP project found out they carry a previously unrecognized mutation in the JAK2 gene, which causes clotting problems and dark spots on the retina. The solution: baby aspirin. “That did involve some research. But the next time somebody comes in with a JAK2 mutation, getting to the point of clinician involvement will be easier and more algorithmic.”Genome sequencing, STAT
When it comes to the $1000 genome, another critical question is whether the price point really matters at all. A desperate patient or family may not mind paying $3000 for a genome, as opposed to $1000. Neither, perhaps, will insurance companies. In some sense, then, the notion of a $1000 genome is arbitrary.
But for researchers like Nusbaum, the cost difference does matter. “The difference between $1000 and $5000 is five times as many samples, and that's five times more [statistical] power,” he says. “Even a two-fold difference in the number of samples is a real difference in how quickly you can get to your answer.”