Scipher Medicine Builds Evidence for Tests Predicting Response to Autoimmune Disease Drugs

Published by GenomeWeb

Scipher Medicine of Boston is hoping to develop molecular signatures that can predict a patient’s response to autoimmune disease drugs, pushing precision medicine into areas other than oncology.

The company, which was founded in 2014 by researchers at Northeastern University, Harvard, and Brigham and Women’s Hospital, uses a proprietary approach for discovering RNA signatures, which it is applying first to predict response to anti-TNF drugs in rheumatoid arthritis and ulcerative colitis.

The firm completed a Series A round led by Khosla Ventures last September, which brings its total funds raised to $10 million. Alif Saleh, Scipher’s CEO, said there are also plans to raise a B round before the summer —money that should help support a planned prospective follow up to the company’s current observational study.

Saleh said that Scipher’s tests rely on the characterization of protein interaction networks in order to identify predictive gene expression signatures organically, from the bottom up, rather using methods common in diagnostic development that compare responders and non-responders, irrespective of whether resulting biomarkers actually reflect disease biology.

In principle, this approach could be applied to a variety of different diseases, including cancer. But Saleh said that starting with autoimmune disease and the anti-TNF drugs was a strategic move.

“As many are aware, the response rates to the anti-TNF therapies — Humira, Remicade and Enbrel — are very, very low. A majority of patients who are prescribed these drugs, up to 70 percent, don’t actually respond to the therapy.” As a result, he said, 70 percent of drugs like Humira, which generated billions of dollars a year in revenue, go to waste.

In the company’s early years, Saleh said, he and colleagues reached out to insurance payors to get feedback on where they thought the greatest need was for precision medicine tools.

“We spent a lot of time with pharma but we felt after a while that pharma is kind of talking out of both sides of their mouth. [On the one hand], they want drug response predictions because it helps them to select patients for clinical trials. On the other hand, it does limit their market. So we asked ‘who really stands to benefit from the ability to predict drug response?’ and the obvious answer is payors and patients,” he explained.

Saleh said the team visited about eight of the 10 largest payors in the US and asked them a “blue sky” question. If they had the ability to predict response to any drug across the full spectrum of human disease, what would it be?

“It was quite remarkable because they all said almost the same thing … don’t give us anything in oncology” Saleh said. “There are [too many] biomarker technologies that are being thrown at us [and] we’re not set up to evaluate what works and what doesn’t work.”

Instead, “literally all of them” wanted predictors of response to anti-TNF, he added.

Other companies have tried to create tools to guide treatment for autoimmune disease, specifically rheumatoid arthritis. A longstanding example is the Vectra DA test, developed by Crescendo Biosciences, which was acquired by Myriad Genetics.

Last year, Myriad said that the test has penetrated only 3 percent of an estimated $3 billion global market. The company had also faced pushback from Medicare contractor Palmetto in 2016, which proposed ending coveragefor the assay based on a lack of prospective clinical utility data.

The Medicare debate over utility evidence for Vectra DA has been limited to the primary use case for the test — in monitoring disease activity — not whether it predicts response to specific drugs, though Crescendo has also tried to collect evidence to that end.

Saleh said that Scipher believes its approach can overcome challenges others have faced. “There are a number of ways of tackling [therapy prediction],” he explained. “You can throw out kind of a traditional machine learning approach to try to find a signature in the gene expression or protein data … [but] rarely do you come up with a signature that works across cohorts and in independent patient groups.

“These genes are supposed to correlate to response or non-response but you can’t really describe why it’s those genes because it’s a black box … So what’s always been lacking is a very clean understanding of biology.”

Scipher’s protein-interaction-based platform has been in development for 10 years, initially in the academic sphere and then, after the company’s inception in 2014, with an eye toward commercialization.

Company co-founders had been involved in the Human Genome Project, Saleh said, emerging somewhat disappointed in the limitations of its outcome. “We spent billions of dollars trying to map out the human genome but we didn’t necessarily get a lot smarter about disease biology. And that really started a conversation here in Boston around [it not being] so much about the genes [but] the proteins.”

Over the last 10 years, Saleh said, the “interactome” research behind what Scipher is now commercializing has resulted in coverage of about “85 percent of all proteins in human cells.”

“As this network grew over time, we became very good at interpreting RNA data using this network and finding signatures that predict response in individual patients to specific therapies,” he added.

Scipher has purposefully not published its technical methods in detail, Saleh said, although members of the company have published academic work that likely underpins some of what the firm is doing. A study by Scipher Cofounder Albert-László Barabási and colleagues in 2017 in Nature, for example, describes a “framework to construct personalized perturbation profiles for individual subjects, identifying the set of genes that are significantly perturbed in each individual.”

According to Barabási and his coauthors, when they examined the heterogeneity of these individual “perturbation profiles” in patients with asthma, Parkinson, and Huntington’s disease, they were able to see a shared pool of sporadically disease-associated genes
such that individuals with statistically significant overlap with this pool had an 80 to 100 percent chance of being diagnosed with the disease.

Scipher also said yesterday that it saw “positive outcomes” in a 150-patient retrospective trial it conducted for its therapy-prediction assay it calls Prism.

This data has not been published in a peer-reviewed journal yet, but according to the company, the test accurately predicted non-response to anti-TNF therapies, including Humira, Enbrel, and Remicade.

According to Saleh, the firm plans to follow up its initial observational study with a larger prospective, interventional validation study, expected to be completed in early 2020.

An interesting side-effect of the disease biology-focused approach Scipher has employed is that the company can also harness it for novel drug discovery.

“The RNA datasets that come out of our blood tests from the non-responders … those datasets are very rich in new protein targets. So what we are also doing is, we’re screening drugs against targets in non-responders … which then feeds an [internal] drug pipeline.”

In terms of future diagnostics, the company already lists ulcerative colitis as a next target on its website, a natural move, considering that it is treated with many of the same drugs as RA. The firm has also said it is investigating the utility of Prism in cardiovascular, respiratory, and neurological diseases.

Saleh said that the company intends to offer tests initially through a centralized lab but hopes eventually to translate its technology and bring it through regulatory approval as distributable kits. Scipher believes, at least, that it will have a leg up in terms of reimbursement, considering its strategy of front-line engagement with payors. “It’s been a challenge in the diagnostic world [to get coverage] and when you talk to [payors], it’s incredibly interesting because they get it. They understand that to some extent, they are a roadblock,” he said.

“But they also are frustrated with companies that are developing tests, spending 20 or 30 million dollars … and then when they feel that they have sufficient evidence for coverage, they go see the payor, as opposed to going to the payor upstream and defining what the product needs to look like, what it needs to do — defining the clinical utility and the financial utility. So we’ve kind of flipped that model.”