Tempus Ramps Up AI-Driven Precision Oncology, Eyes Other Markets

Tempus Ramps Up AI-Driven Precision Oncology, Eyes Other Markets

Dec 03, 2018 | Neil Versel 

CHICAGO (GenomeWeb) – Having raised some $320 million in funding since its launch in 2015, cancer informatics and precision medicine company Tempus continues to ramp up its clinical and molecular data collection and curation to support precision oncology, and is beginning to turn its attention to other disease areas.

In September 2017, Tempus closed a Series C round worth $70 million. That was followed in May by another $80 million infusion of venture capital and then a whopping $110 million Series E round in late August.

The company now has close to 600 employees, mostly in a headquarters facility in Chicago that includes a CLIA- and CAP-certified next-generation sequencing laboratory. Following the most recent fundraise, Forbes valued the company at $2 billion.

Since its inception, Tempus has built a library of molecular and clinical data alongside an operating system to make the data useful for clinicians and researchers alike. The company employs machine learning, next-generation sequencing, and AI-assisted image recognition to enable scientific discovery and help physicians make real-time decisions to guide treatment of their patients.

“Our goal is to basically build the technology systems that would be necessary to usher in precision medicine,” CEO and Founder Eric Lefkofsky said in an interview at Tempus headquarters. “What that means is that there needs to be a technology backbone where you can get structured clinical data, combine that with structured molecular data, [and] combine that with machine learning and artificial intelligence and other analytic capabilities.”

Machine learning relies on structured data. “To bring the power of technology and big data to healthcare, the first step is fixing the data,” Lefkofsky said.

On the biologic side, the Tempus NGS lab now runs four molecular panels: a 595-gene panel called Tempus xT that covers immediately actionable mutations; a 1,714-gene panel called Tempus xO, covering genes linked to cancer; Tempus xE, a whole-exome test that analyzes tumor DNA alongside a normal sample and a whole-RNA transcriptome; and the newest offering, a 77-gene cell-free DNA panel called Tempus xF that launched in September.

Tempus likes to be complete with its sequencing. “When we sequence a patient, we do tumor-normal. We do both germline and somatic,” Lefkofsky said. “We also do full-transcriptomic sequencing on every patient, so we generate lots of data, spanning both genomic and transcriptomic.”

As these are large-panel NGS tests, Tempus mostly deals with patients who have already been diagnosed, particularly those with late-stage, high-risk, or metastatic cancers. “We’re not on the prevention side. We’re squarely focused on when a patient is diagnosed with cancer,” Lefkofsky said.

Lefkofsky acknowledged that insurance reimbursement is lagging for large-scale sequencing of cancer patients, but that is something Tempus is willing to ride out in the short term.

“It’s going to take the payors some time to catch up and realize the value of this data,” he said. “There have been large studies now that have been helpful, where people have looked at very large cohorts of patients that were sequenced and have demonstrated that [20 to 40 percent of] these patients are having durable improvements to progression-free survival or overall survival by virtue of having been sequenced,” Lefkofsky explained.

“This is all moving in the right direction, but our goal is to sequence as many cancer patients as we can, and eventually other disease types, and we are hopeful that eventually insurance will catch up and reimburse this stuff appropriately,” he added.

Tempus stays away from diagnosis, as those almost always come from an anatomic pathologist’s review of slides. “Part of those slides are being sent out to companies like ours to do deep sequencing so that physicians can think about all the different options available to them before they pick a therapeutic path,” Lefkofsky noted.

The company has not released statistics about its patient population, but Lefkofsky said that Tempus expects to sequence more than 100,000 samples in the next 12 months.

Tempus started with a singular focus on cancer. Lefkofsky walked away from day-to-day leadership of Groupon — the company that made him a billionaire — after his wife was treated for breast cancer at a major academic institution. (He remains chairman of Groupon, and Tempus is located in the same sprawling, riverside building as the online daily-deals pioneer.)

“About four years ago, in going through the process with her as a family member, I was perplexed at how little data had permeated her care and just how broken and siloed the underlying data was and how hard it was for physicians to access that data when they wanted to,” Lefkofsky recalled. He previously told GenomeWeb that he was surprised at how few data and technology experts were involved in clinical processes.

In launching Tempus, Lefkofsky put together a team of data scientists, software engineers, bioinformaticians, computational biologists, and research scientists, including President Kevin White, formerly the founding director of the University of Chicago’s Institute for Genomics and Systems Biology.

Lefkofsky also brought in Arul Chinnaiyan, a Howard Hughes Medical Institute investigator and professor of pathology and urology at the University of Michigan, to serve as chief scientific advisor for the company. Tempus uses the Mi-OncoSeq technique developed in Chinnaiyan’s lab, among other sequencing technologies.

Tempus uses Illumina hardware for high-throughput NGS, and several different instrument providers for other testing. Lefkofsky said the company has no intention of making its own sequencing hardware, nor is it looking to get into the direct-to-consumer sequencing business, but it has built its own bioinformatics pipeline, including machine learning and AI.

“What you get out of that is some insight that can be used by a clinician or doctor to make a real-time, data-driven decision,” Lefkofsky said. “It can be used by researchers to do better research. It can be used by pharmaceutical companies to make better drugs. It can be used by payors to pay for the right drugs.”

To date, though, Tempus has mostly dealt with hospitals. Lefkofsky said that the company works with about 75 percent of National Cancer Institute-designated cancer centers in the US, as well as “a few hundred” other hospitals. He claimed that a quarter of all US cancer patients come through Tempus in one way or another.

The company has announced numerous partnerships in precision oncology, notably with the New York University School of Medicine, Vanderbilt-Ingram Cancer Center, the Cleveland Clinic’s Taussig Cancer Institute, the University of California, Davis, Comprehensive Cancer Center, University of Chicago Medicine, and the University of Michigan.

Now, with so much infrastructure in place, the company is expanding beyond its original cancer scope, adding services for diabetes and depression. Lefkofsky said to expect further expansion in 2019.

“Over the last three years, as the company has gotten bigger, we have begun to realize that the same advantages that Tempus is able to bring to the healthcare ecosystem in cancer are applicable in other disease areas,” Lefkofsky said. “We now have a scale advantage,” he added.

Lefkofsky said Tempus picked depression and diabetes because they have some similarities with hereditary cancer.

“There is lots of phenotypic variability, there is lots of molecular data that can be gathered, there is some growing body of evidence, especially on the pharmacogenomics side, that this data adds value,” he explained. “On the other end, there is a large body of drugs that a physician could avail themselves of that are often given to patients in a trial-and-error format.”

In all cases, Tempus is trying to deliver actionable data across the healthcare spectrum.

“Ultimately, when you think about precision medicine, I think it comes in waves. The first wave is using large amounts of clinical and molecular data to help point patient to the right drug at the right time. That will improve the system dramatically,” Lefkofsky said.

“Our hope is that we can use data to empower the entire ecosystem to be data-driven so that instead of clinical trials failing at the last minute, we can help people design trials based upon molecular profiles that will have a much higher degree of success. Instead of a doctor being confronted with a decision and not knowing which path is more appropriate for this patient, [we want] that doctor being able to use data” to pinpoint an appropriate therapy, Lefkofsky said.

“I think you’re going to see a huge shift over the next 10 years or so. We’ll be collecting molecular data a lot. We’ll be analyzing lots of phenotypic and molecular data, and using the output of that analysis to navigate patients down their own kind of unique therapeutic path,” he continued.

“I believe the same thing will exist in cardiovascular disease, neurological disorders, infectious disease, and endocrinology.”