Helping doctors find the best treatment for cancer
By Michelle Ng
Just as a size 9 shoe fits only a certain number of feet, the standard “one-size-fits-all” cancer treatment approach doesn’t work for all patients. In fact, sometimes it actually does more harm than good.
A profound change in cancer care is under way, however. More precise and targeted therapeutic strategies based on new computational tools and machine learning are beginning to emerge, offering the potential to greatly improve care for cancer patients. And The Jackson Laboratory (JAX), through an innovative new data resource called the JAX Clinical Knowledge Base (JAX-CKB), is helping to accelerate progress.
This new resource is connecting doctors with scientists, giving clinical teams around the world access to detailed information that can result in more targeted treatments for patients and less trial and error. In other words, the JAX-CKB is helping oncologists find the most effective treatments for each of their patients, and it’s helping them find the information fast.
An incomplete picture
Oncologists today have access to thousands of cancer drugs, but have little guidance as to which one, or which combination, is the best fit for a given cancer patient.
While some oncologists still examine tumors on slides to determine a patient’s type and stage of cancer, recent advances like new tumor DNA sequencing technologies provide insight into the specific genomic mutations driving a patient’s cancer. This growing wealth of knowledge speaks to exciting advances in the field, while also adding new levels of complexity to the search for the right treatments.
For example, an oncologist might have to parse through thousands of papers to distinguish between clinically significant genomic mutations in a tumor and those that do not contribute to cancer onset and growth. Then there are thousands more papers that may help determine what targeted therapies might treat the patient’s driver mutations. This already arduous process is complicated by the fact that different papers often refer to the same genes by different names. It is easy to get lost in all the literature and can be difficult to come away with the full picture.
This plethora of disorganized information limits oncologists’ ability to prescribe the most personalized, targeted therapies that have the highest likelihood of success. It also prevents tumor boards — specialized teams that design treatment plans for patients — from reviewing more than a few patients each week, despite the needs of many more.
JAX is streamlining this process with the JAX-CKB, a new translational medicine platform that connects clinicians with researchers in real-time, allowing doctors and researchers to spend their time and energy more productively, and help patients more quickly and effectively.
, associate director, genomic market development at JAX, is leading the development of the new platform, which functions like the Google or Bing search engines. The JAX-CKB is a database in which doctors are able to find the most relevant and the most recent information from researchers about a specific gene or therapy.
With search options including genes and treatments, JAX-CKB condenses the vast body of scientific knowledge about specific genes, treatment options, and clinical trials into succinct summaries. These summaries are currently available for 900 genes that have downstream cancer effects and include links to original research.
In essence, the JAX-CKB provides decision support for oncologists answering two questions:
- What genomic mutation is causing this cancer?
- What targeted therapies can treat this tumor, based on its mutational profile
“The goal is to build an informatics platform at JAX that integrates these genetic pieces, works with other groups, and can be offered to the scientific and clinical translational world,” says Mockus.
Machine learning and artificial intelligence
JAX-CKB has received more than 100,000 visits since its launch in 2016. While its current functionalities are making a difference for doctors and patients, the original curation process was astonishingly complex. The database is updated manually, which means that curators have to identify a seemingly endless amount of new papers in order to keep the database up-to-date.
To address this, Mockus and her team have developed a triaging process to help ensure curators are not missing important and relevant scientific papers. This process includes a public-private partnership with Microsoft’s Project Hanover.
Project Hanover creates artificial intelligence to advance precision medicine and is led by Hoifung Poon, Ph.D., who directs the Precision Healthcare machine reading component. Mockus and Poon are working together to create technologies that support doctors and researchers, with immense potential benefits for patients.
Poon believes there is a more effective way to identify the most relevant scientific papers, FDA trials, and clinical trials and curate that information for clinicians. Poon’s team is introducing “machine reading,” which uses intelligent algorithms that can quickly scan through all publications, identify the relevant papers, and highlight their most applicable sections.
This technology would empower curators by reducing their cognitive load and offer a quicker, more efficient method of updating the JAX-CKB. More importantly, it helps ensure that no important information — which ultimately could be the difference between life and death for patients — falls between the cracks.
“The goal is to leave no relevant fact behind,” says Poon.
While Poon’s machine reading system is currently in its pilot stage, he and Mockus look forward to integrating it into JAX-CKB’s curation process.
Mockus’ next steps for the JAX-CKB include rolling out more connections between germline mutations (hereditary genetic alterations in a germ cell such as egg or sperm) and somatic mutations (non-hereditary genetic alterations in a non-germ cell). This additional information about could shed more light on what is causing a patient’s cancer and result in a more targeted therapy for the patient.
She also looks forward to scaling the JAX-CKB by leveraging Microsoft’s machine-learning so that the JAX team can expand its scope in other areas of cancer research, such as the oncobiome.
“By working together and sharing our findings, we may soon have a cure for even the rarest and hardest-to-treat cancers,” Mockus says.