As the tide turns towards personalized medicine based on next-generation sequencing technologies, there will be growing demand for high-performance computing tools that incorporate genome analysis into routine clinical procedures, as well as software that can extract medically relevant data from the millions of sequences in the human genome, Harvard University and Beth Israel Deaconess Medical Center's Mark Boguski told BioInform in an interview this week.
His comments followed last week's announcement that BIDMC, a teaching hospital at Harvard Medical School, formed a two-year partnership with GenomeQuest to develop whole-genome analysis applications for personalized healthcare.
Under the terms of the agreement, GenomeQuest will provide BIDMC with data-management and analysis tools for the hospital's next-generation sequence-based projects in the hope that these tools will enable clinicians to use genomic information to make personalized treatment decisions for their patients.
Boguski said that rather than simply developing applications, GenomeQuest will meet the center's high-performance computing needs without incurring the significant costs associated with setting up a data center.
At the same time, he said, the partnership with GenomeQuest leaves open the option to leverage a "general- and specific-purpose" cloud computing infrastructure for BIDMC's projects, thus eliminating two "huge capital investments" that currently prevent the broad application of clinical genomics.
The Present and the Future
Sequencing vendors and bioinformatics companies have invested significant resources into developing informatics tools that are geared towards the research market, but Boguski said that ultimately, the “real return on investment” will come from the clinical diagnostics market.
Most sequencing companies have realized this point, he said although they haven’t taken full advantage of the opportunity at present. Most bioinformatics companies, on the other hand, “are not thinking about that,” he said.
A reason for this hesitation might be because there isn’t a business model in place that takes into account factors like technical and economic feasibility, HIPAA compliance issues, and economic sustainability, Boguski said. As a result, one facet of BIDMC's partnership with GenomeQuest will be to test what could be a "viable business model" for the space.
While companies like Omicia offer solutions that link whole-genome analysis and clinical practice, GenomeQuest has traditionally provided software-as-a-service geared towards representing, storing, compressing, and indexing genomic sequence data via its GQ-Engine infrastructure, and is a newcomer to the clinical diagnostics market.
Acting CEO Richard Resnick told BioInform recently that the company's data center, which runs thousands of cores, can process 150 whole human genomes at 30X coverage per month and that the firm plans to scale this into the thousands in the near future (BI 09/17/2010).
Pathologists Take the Lead
BIDMC and GenomeQuest plan, among other things, to explore the possibility of creating a cloud computing environment suitable for handling patient data that complies with HIPAA regulations and incorporates electronic medical records, as well as to develop applications that can attach “medically actionable annotation” to genetic data, Boguski said.
As part of their efforts, the partners plan to develop applications that “fit clinical routine procedures [into an] operational infrastructure that exists for other laboratory tests” that are routinely performed by pathologists, particularly for oncology and renal disease.
“Some segment of the medical profession has to be trained to do [clinical genomics]," Boguski explained, adding that the partners opted to work with pathologists over medical geneticists because there are more of them. Indeed there are 17,000 pathologists in the US, but only about 1,000 medical geneticists, according to Boguski.
"At the end of the day, the genotype is another laboratory test … and it fits into the workflow of a clinical pathology operation very easily,” he said.
Including clinical genomics in the pathology workflow is significant, Boguski said, because genomic data will eventually be incorporated across specialties beyond oncology, such as cardiology and neurology. If genomics isn’t made a routine part of laboratory procedure, applying the data across the spectrum of disease could become “unimaginably complex," he said.
"It's not until we begin to practically work through all the steps in the clinical environment, that we are going to know where the roadblocks are and only then we will be able to work on overcoming them," he said.
This integration presents a challenge, however, because specialized platforms for different genetic tests require a wide range of technologies, reagents, and analytical tools.
There are also more informatics-specific challenges. For example, "Are you really going to put a whole genome into a patient's electronic medical record?" Boguski asked. "Probably not."
According to Boguski, molecular diagnostics of the future will rely on a single platform: next-generation sequencing. “What will differentiate one molecular diagnostic test from another will be the algorithms and software filters put on whole-genome or transcriptome data,” he said. As part of its collaboration with BIDMC, GenomeQuest plans to develop such filters, which will sift through the billions of base pairs that make up the human genome and identify which data is useful.
As personalized medicine moves into practice, pathologists ideally won’t have to worry about alignment algorithms or base calling because these capabilities will eventually be “part and parcel” of their instruments, Boguski said.
As an illustration, “the radiologist doesn’t have to know anything about the algorithms by which the images [they see] were reconstructed,” he said. “All of those imaging technologies could not exist without high-performance computing …the computing is built into the equipment.”
While the initial focus of the GenomeQuest partnership will be on cancer and renal disease, which Boguski said is in response to clinicians’ demands, the partners have a much “broader vision" for the future.
For example, he said, newborn babies in the US have to get several genetic tests, which vary from state to state. “At some point, the cost of sequencing is going to be low enough that a whole genome [sequence] … will replace all these genetic tests,” Boguski said.
"So if a newborn gets a complete genomic sequence, when they come in for their regular check-up, that genome, which was done at birth, will be reinterpreted just like when you come in for an annual physical," he explained.
"They won't be resequencing your genome … what they will do is write in orders to reinterpret the genome or update the interpretation of the genome based on the knowledge that’s accrued since that patient was last seen."
Commercial vs. Open Source
Boguski noted that the rise of clinical genomics poses a commercial opportunity for bioinformatics firms that have traditionally been shut out of the genomic research market in favor of open source tools.
As sequencing costs drop, the data interpretation bottleneck worsens and users need "professional systems" — particularly when issues like the regulatory requirements for these systems are taken into account, Boguski said.
In addition, the clinical genome informatics market is currently "not a big enough market to develop and support an open source model," he said. This makes the move of bioinformatics companies into the clinical informatics space "inevitable."
The National Institutes of Health "can't fund every genome center to develop its own software and have it be open source," he said. "Centers who don’t have the expertise or who don’t want to develop their own software are going to have … to use companies like GenomeQuest and others that will spring up to provide these services."