Skip to main content
Premium Trial:

Request an Annual Quote

PointCross Life Sciences Emerges From Stealth to Wrangle Data for Drug Developers


NEW YORK (GenomeWeb) – What do a patient in a clinical trial and an oil well have in common? According to PointCross Life Sciences: plenty, at least from a data management perspective.

Both are subjected to diagnostic testing — what seismic data are to wells, biomarkers are to patients — and both, once selected, are monitored with technologies that produce streams of real-time data.

"When you drill, you get hundreds of sensors that go down the hole," Suresh Madhavan, PointCross CEO, said in an interview. "That exactly what happens on a patient, where you have sensors, you can get EKG, brainwaves, in real time. They're a similar type of problem."

There's one big difference: a patient produces far less data than a well. But accessing and comparing all the data on a patient, especially one in a clinical trial, has been a major challenge. From tissue sample assay data, genotype, phenotype, medical history, drug reference data, and others, it's hard for clinical scientists to get that all in one place and interact with it.

"Though these pools exist today, they're not connected and not searchable together," Madhavan said. That's where PointCross Life Sciences comes in. Programmers there have taken what they've learned from more than a decade of data integration in geology and applied it to biology.

Already working with Bayer and Ipsen, the privately held Bay Area firm is launching its Xbiom platform to help pharma and diagnostics companies get the most out of their clinical datasets.

"It's kind of like a Google search on steroids," Madhavan said. "But it's not just words, it's values. It's actual data. You're not looking for metadata to find the patients in studies, you're looking at the master data that's been collected on the patient and looking for trends in that data."

The potential applications are numerous, such as finding both retrospective and prospective patient cohorts or finding biomarkers for potential companion diagnostics.

It all started with a chance encounter, Madhavan said. In 2010, he found himself listening to someone from Roche talk about its problems with semantic integration of data in non-clinical animal studies.

"They said it was hard to do," Madhavan said. "We showed them how to do it." A representative from Roche's pRed Informatics declined to elaborate on the collaboration but confirmed that it existed.

That success, coupled with the downturn in the oil and gas industry, helped point the firm towards biology, but it wanted to go after the bigger clinical data market.

PointCross is certainly not the first company to approach data integration in pharma. Other firms have attempted to build bespoke databases, but Madhavan said his is the first that can offer streamlined platforms that can be deployed over and over again.

The core technologies in Xbiom are drawn from PointCross' oil and gas products. "It's an ontology engine," Madhavan said. "Unlike relational database management systems, it's a very different way of representing data used in multiple areas. It's designed to bring in data of different types."

PointCross uses a stack of technologies including Hadoop, Spark, and several Apache database and search tools including Cassandra, Lucene/SOLR, and Hive.

It's not just the ability to store data in a single repository, index it, and access it via search queries, but also tools for automating data intake and processing that make PointCross valuable.

"We have a wide variety of semantic data exchanges that allows data coming in from clinical studies or contract research organizations to be brought directly into the system," Madhavan said. What would normally take weeks to collect and analyze manually comes down to buttons on the screen.

Getting, say, a list of patients matching the search parameters from across several different studies is difficult because studies usually don't share the same data model. "What we do is bring in the data, harmonize it, [and] normalize it."

Madhavan offered this example of how a researcher might use Xbiom to find a few dozen patients among a pool of thousands: "You might be interested in asking a question about a particular gene, say, KRAS. You could ask, 'Show me the patients that have a certain type of SNP or CNV at position 12.' You're going from the data and actually finding the people. It's like going to a city and saying, 'Tell me the males who are between 25 and 36 years old and between 5 feet 9 inches and 5 feet 10.5 inches tall.'"

For omics data, Xbiom can also compare results across multiple technologies. For a pharmaceutical trial sponsor using multiple technologies to find biomarkers that correlate to their molecule, PointCross offers a way to compare results evenly between them. The platform also offers information on patient sample availability for re-analysis, including whether the patient has provided consent for that.

The firm is also set up for advances in point-of-care testing and real-time monitoring of patients, thanks to its experience in monitoring drilling wells. "We are waiting to connect up to clients who are doing clinical trials with that kind of [real-time monitoring] technology," Madhavan said.

The company is going to go after pharma companies first, especially in immuno-oncology and cardiology, but it has an eye towards diagnostics. It will often start with a six-month pilot study, running cases using real-world biomarker data. 

So far the response in biotech has been good. "Taking things from another industry, we found we had original ideas that others had not tried," Madhavan said.