CHICAGO (GenomeWeb) – As the name implies, health informatics startup Datavant is all about the data.
"Our belief is that one of the biggest challenges facing data for healthcare is how siloed it is," said Travis May, president and cofounder of the fledgling, San Francisco-based health data integrator and analytics firm. "It's siloed across different pharma companies, it's siloed across providers, it's siloed across payors, it's siloed across universities, it's siloed across analytics companies, etc."
May said that this siloing — certainly not a new problem in healthcare — "holds back a lot of the analytics and a lot of the data-driven decision-making that can transform the way decisions are made within healthcare."
Datavant launched in September as a spinoff from drug company Roivant Sciences, though Roivant announced the new entity in August. Roivant also completed a $1.1 billion equity investment round, led by the SoftBank Vision Fund, in August, and is using some of the money to get Datavant off the ground.
May also made a "significant personal investment" in Datavant, according to Roivant. May said that all employees so far have been made shareholders as well, though Roivant is the majority owner.
Roivant saw the need for an entity like Datavant because the failure rate of clinical trials can be as high as 90 percent, according to May.
"We think that many of those drugs deserve to fail because the drug just intrinsically doesn't work," May said. "But, in many cases, a trial failure is not because the drug intrinsically doesn't work, but because the trial wasn't targeted to the right subpopulation, it wasn't given at the right dosage, there weren't the right inclusion/exclusion criteria, etc."
Datavant wants to organize and help researchers do more with information from sources as disparate as genomic tests and medical billing claims, with an eye toward improving clinical trial design. "Our goal is to organize the data that can optimize those trials," May said.
"The thesis that we have is that there's enough data out there and enough good analytical tools that exist that if you could organize the data that exists, you could roughly double the odds of success for clinical trials," May said.
May said that there are good analytics tools on the market already, so Datavant is leveraging some of those while also creating its own technology. "I think of the data business and the analytics business as adjacent, but different businesses," he said. "Our goal is more to form the data foundation for clinical trials — pulling together the data sets, structuring the datasets, linking the datasets, ensuring that all the data is available."
Datavant then wants to partner with analytics companies and drug developers to interpret the data, adding artificial intelligence and machine learning. "Where the AI and machine learning come in are more around how you structure data," May said.
The startup links a pharma company's internal datasets with "relevant" external data sources, according to May. "In general, most pharma companies, once a trial finishes, the data isn't stored in some central repository and easily repurposed across trials," he noted.
"Step one is to help organize a pharma company's internal datasets and bring structure to that in one central place. The second part is [to] link that to external datasets," May said. External datasets could include information from other sponsors' clinical trial outcomes, electronic health records, genetic tests, billing claims, observations of daily living, and what May called "real-world behavioral activity."
So much information, including published research, claims, and clinical data, are in PDF files, which are not machine-readable. Datavant applies natural-language processing to extract data from this text and apply structure, then store the result in the Clinical Trial Cloud, one of Datavant's core offerings. "There are all sorts of messy data sets that we ultimately want to build and bring structure to," May said.
While drug discovery is Datavant's primary raison d'être, genomics is paramount. The company began 2018 by announcing a series of partnerships with genomics researchers.
One partner is biopharmaceutical company Global Genomics Group (G3). Datavant will offer its AI-driven Clinical Trial Cloud to support G3's G3LOBAL database of whole-genome sequencing, whole-transcriptome sequencing, lipoprotein proteomics, and precision imaging records.
The other deals are with Duke Clinical Research Institute and with AI vendor Verge Genomics. Datavant will provide DCRI with analytics tools and access to 150 different data sources — from genomic tests, drug trials, billing claims, pharmacy records, and EHRs — in hopes of improving design and interpretation of clinical trials. The company will be offering its data sets to Verge for similar purposes.
May said that Datavant has some other partnerships that have not been announced, though some clues can be found on the company's scientific and data science advisory board. This panel includes representatives from PPD Biotech, GlaxoSmithKline, and the MIT Sloan School of Management.
Datavant has just 10 employees now, and is looking to hire data scientists and software engineers. "The core of the team is building a top-tier data science and engineering outfit. [We're] complementing that with the right in-industry experience," May said.
Notable among the team is newly hired Chief Scientific Officer Eric Perakslis, a former chief information officer and chief informatics scientist at the US Food and Drug Administration. Perakslis most recently had been senior vice president for R&D informatics at Takeda Pharmaceuticals and was part of the advisory board announced at the time of Datavant's launch.
"Our goal is to combine the top-tier scientific minds and industry thought leaders with the engineering and data science culture of a San Francisco startup," said May, who actually is brand new to health informatics himself.
Fresh out of college in 2009, May cofounded and later served as CEO of LiveRamp, which performed similar acquisition and analytics tasks on data for the marketing industry. He shepherded that company through a $310 million sale to much larger competitor Acxiom in 2014.
May admitted Datavant was trying to do "ambitious" things in healthcare. "If we can come anywhere close to our goal of doubling the odds of success of clinical trials, it transforms the industry and transforms the economics of drug development," he said.
Right now, about 90 percent of Datavant's activities are around data collection, data acquisition, and forming partnerships, or what May called "trying to expand the whole ecosystem of different data holders." About 15 pilots are underway now.
Datavant wants to work not only with pharma, but also with the EHRs of providers, claims data from payors, and test results from genomics companies, to gather and link their data together.
"We're really pleased with the partnerships we have so far and some of the datasets we've pulled in so far. The most important things for us are getting access to more and more rich datasets that are valuable for clinical trial design interpretation," May said. "Each dataset is more valuable when combined with other datasets."