Bret S. Stetka, MD
Medicine is often a guessing game. Doctors frequently prescribe medications that, while effective across the collective population, may not work for specific individuals. This is why so many clinicians are studying and advocating for personalized medicine; or matching the right patients with the right medications based on their individual biology. Researchers at Scipher Medicine are using cutting-edge technology combined with artificial intelligence to do just that, hoping to improve the treatment of a range of disorders.
By analyzing the network of protein interactions associated with specific diseases and treatment responses, or the “interactome,” they are able to not only predict which medication a patient may or may not respond to but potentially save billions of dollars in wasted healthcare costs due to ineffective prescriptions. Dr. Bret Stetka – a health and science journalist with years of experience covering neuroscience, rheumatology, and genomics – recently interviewed Scipher Chief Technology Officer and Head of Therapeutics, Slava Akmaev, Ph.D., about their technology and aims for the future of clinical care.
In a few words, tell us how Scipher Medicine came to be
This technology comes from years of researchers analyzing the human interactome, which is the network of proteins and protein interactions that influence health, disease, and response to therapy. We obtained the interactome data through a license from Northeastern University, where our co-founder Dr. Albert-László Barabási holds a professorship and spent more than two decades working to map the human interactome. He did so in collaboration with Scipher’s other co-founder Joseph Loscalzo, Chief of Medicine at Brigham and Women’s Hospital
Drs. Barabási and Loscalzo have very different academic backgrounds. How did they come together to found Scipher?
Dr. Barabási is a physicist and has been studying computational biology for years. Dr. Loscalzo is a cardiologist. They brought physics and medicine together and began using computer data modeling techniques that have been used for a wide array of applications. They wanted to see if these approaches could help us better understand biological processes and how protein interactions can predict disease risk and treatment response. The interactions are primarily determined by yeast two-hybrid and mass spectrometry experimental data. In parallel, we have internally, together with the Barabási laboratory, spent many years developing AI and machine learning algorithms that allow us to extract insights from the interactome.
Humans have around 20,000 genes that encode for the proteins that run our bodies. So you’re looking at all 20,000 or so proteins that these genes encode for and how they interact?
Yes. Imagine that you have 20,000 proteins that form this network of physical interactions. Then, take for example, a disease like rheumatoid arthritis, and you ask what are the proteins that are genetically linked to rheumatoid arthritis? Let’s say, maybe it’s a few hundred. We can now map those proteins using machine learning. And then, that connected cluster of proteins becomes what we call a “disease module” that can help us understand which genes and proteins are causing a disease. This will help us identify patient populations based on their molecular phenotypes instead of their observed clinical symptoms. It also allows to identify novel treatment targets.
Can you speak more to how Scipher’s technology could help identify those targets?
These algorithms and disease modules allow us to identify and rank proteins in the order of importance in disease dysregulation. Based on disease module and the network, we can rank every protein in the body – from one to 20,000 – on their potential usefulness as a therapeutic target. We can do it quickly and en masse, in other words for every complex human disease.
What disorders are you currently looking at?
So far, we’ve been looking primarily at autoimmune disorders. We’ve mapped the interactome for rheumatoid arthritis, which has been our biggest focus—also ulcerative colitis and Crohn’s disease. Right now, we’re looking at psoriasis, psoriatic arthritis, lupus, and a few other autoimmune diseases.
How did you decide to start by focusing on RA?
Before I joined Scipher, the leadership team went to some of the largest payers in the United States and asked a fundamental question, “Where is there waste in the healthcare industry? Where’s the biggest pain point?” Unanimously, the response was the anti-TNF therapy – commonly prescribed medications for RA – because only about 30% to 35% of patients respond. TNF-inhibitors are extremely expensive – upwards of $30,000 or $40,000 per treatment course. And despite having high non-response rates, they’re mandated in many formularies as first-line treatment. We used our platform to develop a predictive diagnostic test for anti-TNFs, which launched in the summer of last year. And we’re working on additional tests for many of the conditions I mentioned earlier.
Finally, in addition to developing diagnostic tests, you also foresee Scipher as a therapeutics company, correct?
Yes. We are in diagnostics, but we’re developing therapies and have a total commitment to identify and develop novel treatment targets using our platform. It’s quite an exciting new beginning for us.