$2.2 million awarded to harness revolutions in machine learning for cancer research
Can 3D renderings of pancreatic cancer tumors provide insights that improve the prognosis for one of the deadliest cancer types? Can machine learning help us better predict which patients will benefit from immunotherapies, which currently have a 30-50 percent success rate? And can a smartphone app be a tool in early cancer detection?
These are just some of the questions researchers are looking to answer with funding provided by The Mark Foundation for Cancer Research (MFCR). Details about each project funded in today’s announcement are below.
Most of the research projects in this funding round come out of collaborations formed at a workshop jointly held by MFCR and Carnegie Mellon University in April 2019. Attendees came from many different areas of research including machine learning, computational biology, digital pathology, biomedical engineering, systems biology, and clinical oncology. More on this workshop is here.
“Bringing together scientists from varying disciplines is critical to tackling the toughest challenges in cancer research,” said Ryan Schoenfeld, PhD, Vice President for Scientific Research at The Mark Foundation. “We’re glad to see promising collaborations emerge from our workshop and to be the first foundation with such a robust portfolio at the intersection of AI and cancer research. We’re just at the beginning stage of harnessing the incredible power of machine learning in the fight against cancer, and we’re excited to see where these projects take us.”
Michael Schatz, PhD of Johns Hopkins University and Eliezer Van Allen, MD of the Dana-Farber Cancer Institute joined forces in one such collaboration. Michael and Eli will use an advanced genetic sequencing technology specifically designed to detect structural variations (SVs), a class of DNA mutations that are not typically detected by standard DNA sequencing technologies. Their approach could uncover harmful SVs in families that have high rates of unexplained cancer.
“We are incredibly grateful to The Mark Foundation for supporting this research and promoting our new collaboration,” said Schatz. “The new technologies will let us examine these families with unprecedented resolution, and we expect to find tens of thousands of variants per patient that have never been observed before. Together, we will evaluate this new class of cancer risk variants and aim to improve cancer prevention, improve cancer therapy, and ultimately save lives.”
Researchers from the following institutions/organizations have received funding: 4YouandMe; Carnegie Mellon University; Children’s Hospital of Pennsylvania; Dana-Farber Cancer Institute; Institute for Systems Biology; Johns Hopkins University; SickKids Research Institute/University of Toronto; University of California, Davis; and University of Pennsylvania.
Descriptions of the AI/cancer research project funding announced today are below.
Title: Using Smartphones and Wearables for Early Detection of Central Nervous System (CNS) Tumors
Researchers: Stephen Friend, MD, PhD, 4YouandMe (Principal); Anna Goldenberg, PhD, The Hospital for Sick Children (SickKids) Research Institute / University of Toronto; Marzyeh Ghassemi, PhD, Vector Institute/University of Toronto; Luca Foschini, PhD, Evidation; et al
Description: For the 80,000 patients diagnosed with brain and central nervous system cancers each year, most of the provision of care rests on subjective, discontinuous data collected at discrete time intervals. To address this limitation, a coalition of researchers led by Dr. Stephen Friend of 4YouandMe will test the feasibility of using smart devices and health tracking apps to detect symptoms in brain cancer patients. They will conduct an observational study of 100 high-risk patients and use the wealth of data collected to develop a model of disease progression. Using each patient’s own smartphone and small wearable devices in conjunction with health apps that collect, compile, and transmit data directly to clinicians, the coalition will be able to semi-continuously monitor changes in everything from gait to mental health to sleep. They expect analysis of the data will reveal patterns of symptoms that are missed by traditional patient monitoring. These insights could enable doctors to detect tumors early, before symptoms become noticeably severe, and to closely monitor and adjust treatments in patients with advanced disease.
Title: Using Blood Biomarkers to Aid App-Based Cancer Monitoring
Researcher: James Heath, PhD, Institute for Systems Biology
Description: The data collected by smartphones and wearables will be even more powerful when combined with information about patients’ molecular markers of cancer. Toward this goal, researchers in the Heath lab at the Institute for Systems Biology are collecting and analyzing in-depth measurements of blood biomarkers in the patients monitored for symptom detection using smart technologies by Stephen Friend and colleagues. Working together, both teams expect that tracking the trajectory of these physical markers will allow them to calibrate the output from devices and lead to algorithms that optimally connect both types of data for early screening and surveillance of high-risk patients.
Title: Using Artificial Intelligence to Predict Interactions between Immune and Tumor Cells
Researcher: Alexander Baras, MD, PhD, Johns Hopkins University
Description: Understanding how immune cells and tumor cells will interact is critically important for forecasting patient outcomes for immunotherapy. The Baras lab at Johns Hopkins University is designing artificial intelligence-based algorithms that can identify how mutations are most likely to be presented by tumor cells given a patient’s unique genetic make-up and predict how well immune cells will be able to detect them. These algorithms may lead to improved prognoses of individual patients and a better understanding of how each patient will respond to treatment.
Title: Artificial Intelligence Assisted MRI Screening for Pediatric Cancers
Researchers: Anna Goldenberg, PhD, SickKids Research Institute/University of Toronto (Principal); Andrea Doria, MD, Hospital for Sick Children; Casey S. Greene, PhD, University of Pennsylvania; L. J. States, MD, Children’s Hospital of Pennsylvania
Description: Whole body magnetic resonance imaging (wbMRI) is an important tool for screening children with a genetic predisposition for cancer. However, the images can be difficult to interpret, and early stage cancers are often missed. The Goldenberg lab at the SickKids Research Institute in Toronto, in collaboration with researchers at the University of Pennsylvania and Children’s Hospital of Pennsylvania, is developing an AI-based tool that improves wbMRI screening for at-risk children and overcomes current limitations to using AI in the pediatric cancer setting. They expect that this tool will enable more sensitive early detection of pediatric cancers at medical institutions worldwide.
Title: Detailed, Automated 3D Imaging of Pancreatic Cancers and Precancers
Researchers: Richard Levenson, MD, University of California, Davis (Co-Principal); Farzad Fereidouni, PhD, University of California, Davis (Co-Principal); Ralph Hruban, MD, Johns Hopkins University; Denis Wirtz, PhD, Johns Hopkins University; Laura Wood, MD, PhD, Johns Hopkins University; Pei-Hsun Wu, PhD, Johns Hopkins University
Description: Pancreatic cancer is an aggressive disease that readily spreads to other organs, particularly the liver. Researchers in the Fereidouni and Levenson labs at UC Davis have teamed up with leading experts in pancreatic cancer at Johns Hopkins University to develop a fully automated 3D microscopy technique that can be used to image pancreatic tumors and nearby blood vessels, enabling closer study of the invasion of small veins into tumors. They expect that this will lead to a better understanding of the poor prognosis for pancreatic cancer patients and provide cancer researchers with a precise and affordable tool to study tumor anatomy.
Title: Detecting Novel Cancer Mutations That Change the Genome’s 3D Structure
Researchers: Jian Ma, PhD, Carnegie Mellon University (Principal); Eliezer Van Allen, MD, and Felix Dietlein, MD, PhD, Dana-Farber Cancer Institute
Description: Mutations in the region of the genome that are important for the 3D structure of chromosomes are likely critical for cancer growth, but nevertheless poorly understood. Researchers in the Ma lab at Carnegie Mellon University and the Van Allen lab at the Dana Farber Cancer Institute are developing machine learning algorithms trained to detect mutations that are likely to affect the genome’s 3D structure. This technology will pave the way to a new understanding of the influence of these mutations on tumor proliferation, clinical outcomes, and patient responses to existing and emerging therapeutics.
Title: New Imaging Methods for Identifying Structural Differences in Cancer Cells
Researchers: Robert F. Murphy, PhD, Carnegie Mellon University (Principal); Min Xu, PhD, Carnegie Mellon University; Yi-Wei Chang, PhD, University of Pennsylvania
Description: Tumor heterogeneity presents a major challenge in predicting outcomes and appropriate therapies for cancer patients. Different tumor cells may have distinct physical characteristics. Electron cryotomography (ECT) has unrivaled power to produce 3D images of cells and tissues at unprecedented spatial resolution. Researchers at Carnegie Mellon University led by Robert F. Murphy and Min Xu and at the University of Pennsylvania led by Yi-Wei Chang are developing machine learning methods that can be used in combination with ECT to distinguish differences between cancer cell types and determine whether those differences can be correlated with disease progression and treatment outcomes.
Title: Advanced DNA Sequencing for Uncovering Novel Inheritable Carcinogenic Mutations
Researchers: Michael Schatz, PhD, Johns Hopkins University (Principal); Eliezer Van Allen, MD, Dana-Farber Cancer Institute
Description: Many heritable cancers have no known genetic causes. This is in part because standard DNA sequencing technologies do not normally detect an entire class of mutations called structural variations (SVs). Researchers in the Schatz lab at Johns Hopkins University and the Van Allen lab at the Dana-Farber Cancer Institute have developed a genetic sequencing technology specifically designed to detect SVs and are using it to find harmful SVs in a cohort of families that have high rates of unexplained cancer. They expect this will lead to an immediate improvement in screening and diagnostics for heritable cancers, allowing doctors to intervene earlier by identifying those at heightened risk for disease.