COMPUTING CARE

COMPUTING CARE

NEWS

BY ERIN MCCALLISTER, SENIOR EDITOR

Tempus Labs Inc. has created a machine learning tool that can help physicians make better treatment decisions by comparing a cancer patient’s genetic and clinical profile to those of thousands of other patients with known treatment outcomes.

The tool supplements molecular profiling conducted in the company’s CLIA lab by placing the results in the context of a vast set of data — including genetic information, pathology reports, physician notes and treatment outcomes — that are otherwise unavailable to treating oncologists.

“No other single company out there is bringing this all together to help physicians make data-driven treatment decisions,” President Kevin White told BioCentury.

CEO and co-founder Eric Lefkofsky told BioCentury Tempus was borne of frustration from navigating treatment options when his wife was diagnosed with breast cancer in 2014. “I was perplexed at how little data permeated care,” he said. “It became apparent that the only way to usher in precision medicine was to fix the underlying data infrastructure in cancer.”

Lefkofsky, who previously founded the online discount company Groupon Inc., said his goal was to create a tool capable of integrating an array of data types from past and present patients and using the information to craft personalized treatment plans for new patients.

Tempus started by aggregating public data, then expanded its collection through collaborations with academic institutions and research centers such as the National Cancer Institute (NCI).

Tempus is not disclosing the size of its database, but Lefkofsky said it has cohorts for a variety of tumor types that are 10-20 times bigger than those available through NCI’s Cancer Genome Atlas. The atlas contains 2.5 petabytes of data from more than 11,000 patients.

Lefkofsky said Tempus designed its platform to find relevant pieces of information in unstructured physician notes, such as patient demographics and treatment history, and integrate them with structured molecular and clinical data. It also updates in real time as oncologists submit new information about their patients’ responses to therapy and side effects, he said.

To create a molecular profile for a new patient, Tempus analyzes samples of tumor and healthy tissue in its CLIA lab.

The biotech has four different tests doctors can choose from. xT is a panel of 595 genes relevant to diagnosis, prognosis and therapeutic targeting of cancer. xO looks at 1,700 somatic and germline DNA and RNA sequences. The company also offers its xE whole exome sequencing test and xG whole genome sequencing. The company plans to launch a liquid biopsy screening test in 1Q18.

After feeding the molecular profile into its platform, Tempus delivers a report to the physician in two to three weeks ranking available treatments, including those in clinical testing, by how strongly it predicts the patient will respond.

The report also contains treatment and outcome information from de-identified patients it finds most molecularly and clinically comparable to the individual.

Traditional molecular genetics companies validate their tests once, using data from select clinical cohorts, and do not continuously update their algorithms or factor unstructured data into their analyses. White said Tempus is the only company that provides information from matched comparator patients.

Cancer genomic profiling company Foundation Medicine Inc. does have a database that contains structured and unstructured data from various sources. It is using that database purely for research, drug discovery partnerships and stratifying patients for clinical studies.

Tempus sells its tool to medical institutions as a laboratory-developed test. Lefkofsky said the biotech has had mixed success at obtaining reimbursement from commercial payers. It has not yet decided whether it will seek approval for the tool as a molecular diagnostic, which could improve reimbursement.

The company also generates revenue through research partnerships. Its platform will be used to match patients to treatments in The Pancreatic Cancer Action Network’s Precision Promise platform trial. The adaptive study will start in 2018 and test immunotherapies, DNA damage repair agents and stromal disruption treatments.

COMPANY PROFILE

Tempus Labs Inc.

Chicago, Ill.

Technology: Machine learning platform to match patients to therapies

Disease focus: Cancer

Clinical status: Marketed

Founded: 2015 by Eric Lefkofsky

University collaborators: Mayo Clinic, Cleveland Clinic, OHSU Knight Cancer Institute, UC Davis Comprehensive Cancer Center, University of Virginia, Rush University Medical Center, Moores Cancer Center at UC San Diego Health, University of Chicago, University of Michigan, Robert H. Lurie Comprehensive Cancer Center of Northwestern University, Abramson Cancer Center of the University of Pennsylvania, Duke Cancer Institute, University Hospitals Seidman Cancer Center

Corporate partners: None

Number of employees: 200

Funds raised: $130 million

Investors: NEA, Revolution Growth, Eric Lefkofsky, Brad Keywell, John Doerr

CEO: Eric Lefkofsky

Patents: None issued

COMPANIES AND INSTITUTIONS MENTIONED

Foundation Medicine Inc. (NASDAQ:FMI), Cambridge, Mass.

Groupon Inc. (NASDAQ:GRPN), Chicago, Ill.

National Cancer Institute (NCI), Bethesda, Md.

Pancreatic Cancer Action Network, Manhattan Beach, Calif.

Tempus Labs Inc., Chicago, Ill.

REFERENCES

Cukier-Meisner, E. “Yes we PanCAN.” BioCentury (2016)