In 2017, the American Cancer Society awarded $112 million for cancer research.[1] The National Institute of Health FY17 budget for cancer research was nearly $6 billion.[2]
Pharma companies are pouring billions of R&D dollars into developing new treatments for cancer and are looking for the best ways to maximize those investments. One of the, most effective research tools available and one increasingly leveraged by biopharma is artificial intelligence (AI), used to help unlock the complex mechanisms of cancer.
Oncology is an enormously complicated field, so it’s no surprise that researchers are leading the charge in incorporating AI into their discovery and development process. Researchers and data scientist are using the power of AI to help advance cancer research in many ways, from uncovering the mix of biomarkers that can pinpoint a patient’s reaction to a treatment to identifying the best candidates for clinical trials.
Why AI?
The era of single disease markers is over. Treating cancer requires taking a holistic view of all possible patient biomarkers. Solving the puzzle is particularly arduous because our understanding of cancer heterogeneity is increasing, and personalized treatments are becoming ever more imperative. Parsing available data to find proper biomarkers and treatment is critical because “a potentially effective therapy could be overlooked because the indicator for that drug [such as a specific gene mutation] wasn’t found in the biopsy,” says Dr. Lowe of Memorial Sloan Kettering Cancer Center.[4]
To uncover these insights critical to selecting the right treatment for the right patient, researchers need to be able to synthesize massive amounts of data concurrently – something traditional research methods make difficult. The growing mountain of available data, including genetic and tumor markers, combined with the dramatic increase in computing power paves the way for AI to be leveraged and enables researchers to discover insights faster, accelerating the drug development and clinical trial process.
A Multiple Myeloma Research Foundation research project provides an example of how AI can uncover insights undetectable through traditional methods. By creating a comprehensive causal disease model for multiple myeloma, using the GNS REFS platform, the Foundation and GNS identified the CHEK1 gene as a causal driver and predictive biomarker of patient’s response to stem cell treatment. Patients with low levels of CHEK1 were likely to receive a 22-month progression free survival benefit from stem cell transplantation, while those which high levels did not see a significant benefit.
Why now?
Several efforts are currently underway to leverage digitized and standardized cancer related data. These initiatives are primarily geared toward teaching AI algorithms to recognize tumors or cancerous cells faster than humans can. The impetuous is understandable – cancer treatment is already a long process and timeliness can weigh heavily on long term outcomes. With the growth of available data, there is more source information on which to train the algorithms, a critical function to ensure accurate findings.
This aspect is a top concern for companies as well as the FDA. The government agency’s Oncology Center of Excellence is working with Friends of Cancer and the National Cancer Institute to codify reference standards for assessing tumor mutational burden and identify which patients are most likely to respond to immunotherapy.[5] This partnership, coupled with recent announcements from the FDA on establishing a clearer space for AI in healthcare, assert that the agency’s recognition of the power of AI for cancer.
From a financial perspective, the American government as a whole has committed significant funding to fight cancer. In the 2016 State of the Union address, former President Obama launched the Cancer Moonshot initiative authorizing $1.8 billion in spending over seven years for cancer research. With the end of that spending not even on the horizon, President Trump called for a $500 million funding increase in medical research specifically for childhood cancers over the next decade. The financial incentive for continued work in the field is undeniable and is yet another reason to activate innovative technologies such as machine learning and artificial intelligence.
Where it works?
AI has been crucial to the growth of immunotherapy – a course of treatment that uses substances that occur naturally in the body to enhance the effectiveness of the immune system – and is increasingly being used in the battle against cancer. Known as immune-oncology when applied to cancer, these new treatment methods are being combined with traditional therapies like chemotherapy, radiation, and surgery.
AI helps determine the right combination of therapies based on an individual biology. To make the right decision, it’s essential to understand why an individual patient is responding to a certain treatment or how he or she is likely to respond based on genetics. AI likely holds the key to answering many of those questions.
The diversity of disease is becoming more apparent and our ability to treat patients needs to catch up. When it comes to oncology, researchers have been actively incorporating AI into the discovery and development process, but it’s not only a question of cancer. Researchers in other disease states need to follow suit to deliver on the promise of precision medicine.
[1] Current Investment by Areas of Research, Extramural and Intramural Investment by Area 2017, American Cancer Society website.
[2] HHS FY2017 Budget in Brief NIH, NIH Budget Overview, Department of Health and Human Services website.
[3] Is Biopharma Investing Too Much in Cancer R&D, by John LaMattina, Forbes, April 24, 2018.
[4] https://www.mskcc.org/blog/what-tumor-heterogeneity
[5] Gottlieb: FDA to expand real-world data infrastructure to enhance AI capabilities, Cancer Letter, February 1, 2019.
