Genomic Profiling for Precision Medicine: Interview with David Spetzler, Caris Life Sciences

Caris Life Sciences, a Dallas-based innovator in molecular science focused on fulfilling the promise of precision medicine, has developed the MI Genomic Profiling Similarity (GPS) score to compare the molecular characteristics of specific tumors against those in the Caris database. This allows clinicians to identify the molecular subtype of their patients’ tumors and guide personalized treatment. The system is driven by machine learning algorithms and is envisaged as being particularly useful in guiding the treatment of cancers in cases where there is ambiguity about the tissue of origin and in other atypical or difficult to treat cancers.

Personalized medicine holds enormous potential in improving patient outcomes, but has been limited by the complexity of contextualizing the molecular or genetic signatures of diseased tissues, such as tumors. While modern experimental techniques can rapidly generate enormous amounts of molecular data for a specific patient sample, these data are meaningless unless interpreted correctly and compared with other molecular signatures to provide appropriate context.

Artificial intelligence techniques are well-suited to this task, and AI solutions may be the key to unlocking the potential of personalized medicine. In this vein, Caris Life Sciences has used machine learning techniques to develop the MI GPS score that allows clinicians to compare the molecular characteristics of their patients’ tumors with thousands of others in the Caris database.

The system exploits a substantial database of molecular data with associated clinical outcomes, and machine learning algorithms help to identify molecular patterns associated with specific cancer subtypes and treatment outcomes. Using Caris Molecular Intelligence, the company is able to assess DNA across a 592-DNA gene panel, gene fusions, RNA splice variants, gene expression and proteins.

Caris has reported that the MI GPS score demonstrated over 95% accuracy in classifying tumors from a total of 55,780 samples, and indicated a tissue of origin for the vast majority of tested samples for which a tissue of origin was uncertain or unknown.

Medgadget had the opportunity to talk to David Spetzler, President and Chief Scientific Officer at Caris Life Sciences, about the company’s advances in this field.

Conn Hastings, Medgadget: How did you get interested and involved in precision medicine and artificial intelligence?

David Spetzler, Caris Life Sciences: I have always been interested in science. I worked in a neurobiology lab when I was 14 and, before Caris, I was a member of the research faculty at Arizona State University, where I studied mathematics and molecular cellular biology with a focus on the application of novel analytic methods to complex problems. I joined Caris in 2009 as a bioinformatics scientist to analyze the incredible data set that Caris was generating. The combination of mathematics and biology have created incredible opportunities to advance our understanding of cancer and AI is a natural extension of that, enabled by advances in computing hardware.

The best way to help people with cancer is to find it early enough that it’s curable. To that end, I am passionate about understanding the molecular drivers of disease so that we can identify optimal treatment strategies for each patient, and I look forward to continuing to develop early cancer detection assays, discover novel drug targets and characterize protein differences in each patient’s tumor. Precision medicine improves the lives of people living with cancer by finding therapies more likely to work and reducing the risks of patients receiving unnecessary, expensive therapies that don’t impact the course of their disease. This is what drives my colleagues and me at Caris.

Medgadget: What are the advantages of precision medicine?

Spetzler: Precision medicine and next generation sequencing can help to match patients to the appropriate therapies and can determine if a tumor will be more likely to respond to a treatment before it’s prescribed, allowing for more personalized treatments and enabling patients to realize the fullest potential of targeted cancer therapies.

Personalized medicine can also help to improve health outcomes and lessen spending by patients, hospitals and insurance companies, particularly since the average cost of treatment with modern cancer therapies averages about $250,000 per patient. Next generation sequencing allows Caris to rapidly examine and more broadly detect DNA mutations, copy number variations and gene fusions across the genome and potentially change the way clinicians approach treatment for their patients.

Medgadget: How close are we to using precision medicine routinely for every patient?

Spetzler: Right now, treating oncologists routinely order molecular profiling for only 15% of their patients, meaning that 85% are not receiving the testing they need that would make their treatment consistent with National Comprehensive Cancer Network (NCCN) guidelines. Getting patients on the appropriate treatments is a critically important step since it directly correlates to improvements in overall survival and can have a significant impact on the cost of their therapy.

So, while these tests are readily available and recommended for routine use by NCCN, utilization is still a challenge. Caris is working diligently with leading institutions to make these tests available for patients living with a cancer diagnosis, and we are hopeful that uptake continues to increase so that more people are receiving the best possible treatments as early as possible in the course of their disease.

Medgadget: So, what is the MI GPS score, and how does it work? How has it helped with patient treatment so far?

Spetzler: MI GPS Score is an AI-driven tumor type biology similarity score that uses more than 6,500 mathematical models in the machine learning algorithm to compare molecular characteristics of a patient’s tumor against Caris’ extensive database to provide new insights into the molecular subtype of cancer of unknown primary (CUP) cases, atypical clinical presentation cases and other difficult to treat cancer cases to help guide treatment decisions.

If an oncologist is unable to identify the tumor type or genetic lineage of a patient’s cancer, they can send a tumor sample to Caris for evaluation by our team of expert pathologists. The reports our pathologists develop can be used by physicians to make more informed treatment choices.

Medgadget: Please explain the role of artificial intelligence in the development of the MI GPS score.

Spetzler: The behavior of cells is driven by changes in their molecular system, which is one of, if not the most, complex system there is, with literally hundreds of trillions of different interactions occurring per second. The complexity of the system is too great for humans to decipher alone, hence the need for AI and machine learning systems. Using the power of DEAN (Deliberation Analytics) artificial intelligence and machine learning, Caris can provide oncologists with thorough genomic and molecular analysis classification. The combination of AI and human intelligence provides the most comprehensive analysis available today to characterize a patient’s tumor and support treatment decisions.

Medgadget: How do you see this type of technology developing in the next ten years?

Spetzler: Caris and its partners will continue to accumulate data that informs, expands and refines our existing algorithms so that as the database grows, so do the accuracy and precision of the test results. MI GPS Score is the first of many Caris Molecular Artificial Intelligence offerings that will advance our understanding of cancer and enable better patient care.

We will continue to work to improve healthcare for everyone using our unique and transformative platforms to help patients with cancer and other complex diseases.



Source