NEW YORK (GenomeWeb) – The University of Michigan will invest roughly a quarter of a planned $100 million investment in data science activities at the institution in existing and new genomics and personalized medicine research efforts.
The university said this week that the new Data Science Initiative will help students and faculty across the institution involved in transportation research, health sciences research, learning analytics, and social science research to gather, store, search, and analyze large quantities of complex information.
As part of the initiative, the university has established the Michigan Institute for Data Science (MIDAS) to lead activities across all four target research areas. The institute will be co-directed by Alfred Hero, a professor of engineering, and Brian Athey, a professor and chair of the department of computational medicine and bioinformatics. In addition, UM plans to hire 35 new faculty from various disciplines over the next four to five years; expand its research computing capacity; and provide additional data management, storage, analytics, and training resources as part of the initiative.
In the context of biomedical research and personalized medicine, the initiative will seek to equip researchers with tools that let them combine large quantities of next-generation sequence data with information from medical histories and other sources, and to use this information in tandem to assess individuals' risk for diseases and to select more effective personalized therapies.
The new institute along with the planned financial investments will support a number of ongoing projects at UM covering a wider range of disease areas and spanning multiple centers and schools, Athey told GenomeWeb. The emphasis won't be solely on integrating data from the various omics divisions — genomics, epigenomics, proteomics, and so on — but also on encouraging multi-center collaborations focused on combining and exploring omics information in the context of broader, more heterogeneous data streams from other sources including demographics, environmental exposures, and socio-economics.
"Our vision is to go from the basic molecular through the biomedical application all the way out to outcomes," Athey said. "Our role as the co-directors of the initiative is to build off of this strong base of research across our schools and colleges and make investments where we can really make a difference not just [in] doing genomics but adding genomics to other kinds of measurements."
Existing projects that MIDAS would support include the Michigan Genomics initiative, which is currently sequencing and analyzing whole exomes from about 125,000 samples from consented patients in the UM health system, Athey said. So far, that project has sequenced about 55,000 specimens.
Funds would also support the Michigan Oncology Sequencing Center (MI-ONCOSEQ) research study, he said, which aims to use patients' somatic and germline sequence information to personalize treatments and also explores psychosocial and ethical issues associated with disclosing genomic results to patients and clinicians. Other projects at the university focus on developing methodologies for clinical trials, including using wearable devices to track patients and correlating collected information to different phenotypes, as well as on measuring physiological signals in patients in intensive care units to identify clinical decision support metrics.
UM is planning an inaugural symposium to mark the launch of its Data Science Initiative on Oct. 6, and MIDAS has already begun recruiting candidates for faculty positions in the center. Of the 35 new hires, Athey expects roughly six to 10 recruits across multiple departments will likely be involved in research areas that bear on personalized medicine and health.
MIDAS is also looking into infrastructure investments to support data science activities on both the hardware and software side, Hero said. That includes tools for visualizing high-dimensional heterogeneous datasets that could help researchers better locate correlations between genomic, behavioral, cognitive, and other kinds of data, he said. There will also be investments in computational hardware and software frameworks to enable researchers to run the requisite algorithms they need to make sense of complex datasets.
"There's a lot of machine learning [and] a lot of high-level computational modeling involved, and of course to host that, you need a system which is capable of analyzing very large datasets," Hero said. To that end, "We are installing high-throughput Hadoop ... enabled infrastructure that will create both an access point for researchers to these various types of data and also permit hybrid computational algorithms that combine everything from epidemiological susceptibility, infection recovery models, down to molecular-level kinetic models [to run] together in one computational engine."
MIDAS is also exploring industry partnerships and has been discussing its data science initiative with some key companies.
"The areas of precision medicine and personalized medicine and healthcare have a lot of leading companies that we need to be working with," Athey said. As such "We are very open to building partnerships not only with companies like Illumina and Life Technologies but companies that you might not think of immediately, [such as] Northrop Grumman, which has a large footprint in the defense and healthcare industry."