High Performance Computing in Cancer Research

June 9, 2017 Arin Karimian

Advance the use of high-performance computing in cancer research

Eric Stahlberg is the Director of High-Performance Computing at the Leidos-operated Frederick National Laboratory for Cancer Research (FNLCR). In February, the federal technology publication FCW named Dr. Stahlberg to its 2017 Federal 100 list for his work in predictive oncology. Stahlberg’s efforts led to a successful collaboration between the National Cancer Institute (NCI) and the Department of Energy (DOE) that has helped advance the use of high-performance computing in cancer research.

Leidos.com’s Arin Karimian recently spoke to Stahlberg about the Fed 100, his work at FNLCR, and the future of cancer research. Here’s their conversation:

What does being named to the Fed 100 mean to you? 

When I received it, I was quite surprised. Being recognized for doing something that I enjoy and am passionate about is a nice confirmation that my efforts are making a contribution in an area that matters. Likely the biggest take-home point is that the award confirms not only the work I do, but also that the work of the many others contributing to the effort, is something appreciated, has an impact, and that it does matter. 

What is predictive oncology?

Predictive oncology is a concept that aims to take available knowledge, data and insight about cancer, and employ computational approaches to develop a prediction of what will happen. The concept applies across multiple levels in cancer, moving beyond describing what one currently observes, to actually enable a prediction of what will be seen or become of the cancer. It's about acquiring enough information about the disease so that the individuals studying and treating cancer can have the ability to develop hypotheses and project what will happen and take an appropriate course of action. This is the essence of what is emerging as predictive oncology.

The longer-range goal is to reach prescriptive oncology, where enough information would be available — whether it’s from analytics, existing information, models or simulations — that supports the decision of a scientist or oncologist. When predictive oncology becomes prescriptive, there is high confidence in the predictions.

So, is it accurate to say that the next advancement or breakthrough in the space is to get to that prescriptive model?

I would say that is the longer-term aim. With such a complex disease, being able to accurately predict is a tremendous challenge. Hopefully, predictive oncology is moving things forward to what we envision would be a prescriptive model.

The Precision Medicine, Strategic Computing, and Beau Biden Cancer Moonshot℠ Initiatives all launched in the last two years. How does predictive oncology fit into those national initiatives?

When the Precision Medicine Initiative started in Jan. 2015, it provided the basis for motivating the development of information for precision medicine — the ability to gather enough insight on the genomes of individuals. With this information, one can better and more precisely characterize individual patient cohorts and small sub-populations of patients so that you have the ability to more precisely target treatments or approaches to a particular subset of patients. This is what enabled the rise of predictive oncology. 

The importance of the National Strategic Computing Initiative, which was also announced in 2015, was that it provided the basis and motivation for doing cross-agency work in strategic computing. This initiative is a critical piece in this whole story because the NSCI allowed and enabled the NCI to work closely with the DOE and leverage the DOE’s years of experience developing predictive models and employing computational science on extremely complex systems. In the collaboration, we have worked together to leverage that knowledge, experience and capability, and applied it to cancer research.

The Cancer Moonshot capitalized on these initiatives, providing a broad base of support. The moonshot doesn’t just focus on the narrow range of predictive oncology. Instead, the Moonshot is much broader in terms of developing networks of information, understanding specific cancers, developing new technologies, and really working together to understand some of the key challenges in the cancer field. All three initiatives have worked well together to move things forward.

Let’s go back to cross-agency collaboration. How did the connection to DOE begin? 

One of the most important things to recognize about this collaboration is the support that so many people provided along the way as the idea was incubating and developing. The support was tremendous and it’s actually what made it all possible. As I’ve told some of my colleagues, you can’t imagine how seamlessly things fell into place. You couldn't write a script that would have been better in terms of this being just the right time to push this effort forward.

To answer your question, I have a long history of working with the DOE, all the way back to being a graduate student at Ohio State University and having the opportunity to work with researchers at Argonne National Lab. While a post-doc at Argonne, we were really pushing the limits of high-performance computing of the time, with new types of applications in chemistry. The experience at Argonne played a big part in my career to pursue opportunities that integrate high-performance computing and computation to push the limits and solve some of the most challenging problems.

The unexpected origins of this effort began in 2010. I moved to the Frederick National Lab to help start a bioinformatics core supporting NCI’s Center for Cancer Research. While directing the bioinformatics core, I also had a personal situation develop that served as motivation. 

Like many families, my family was touched by cancer with the diagnosis of my sister with cancer. Seeing my sister’s experience, and unfortunately knowing many cancer patients and survivors personally, I felt compelled to take my experiences in HPC, bioinformatics, and computational science, and just see if there was a way to make a difference for cancer patients. With an aim of impacting patients, I began working to convince others that we should try to accelerate cancer research by raising the level of commitment and utilization of high-performance computing. Through contacts, we learned of an initiative led by Lawrence Livermore National Lab building interest in biological applications of advanced strategic computing that really moved things ahead. It is so important to recognize all of those along the way who became supporters and champions across Frederick National Lab, the NCI, the DOE, and the DOE labs, all of whom have made the collaborative effort possible.

National Cancer Institute & Department of Energy Collaborations

What can this successful collaboration teach other agencies? What do you hope they can learn from it?

This experience has taught me that there are many situations where agencies will work well together. I think in our case, having a challenging problem people can personally relate to and then identifying an opportunity where they can have a role and an impact, helped make it possible. 

There are several lessons to take away. At an agency level, it's really important to look outside of your domain for those that have extra expertise in the areas where you're looking to grow, develop, and identify opportunities. One also needs to be prepared that it will take work to get started, bringing interests together. Yet with the right problem and a common motivation, working through early challenges leads to incredible opportunities and a cohesive team effort.

A new nationwide study in the journal Cancer finds that counties with the poorest environmental factors have higher cancer rates. How should findings like this drive cancer research? 

These types of studies are likely to be enhanced in the scope of the pilot focusing on cancer surveillance. This pilot aims to explore the requirements to broadly and more comprehensively integrate cancer patient information together with environmental information, providing better insight into factors that may be leading to such disparities. We anticipate the capabilities being piloted in the NCI-DOE collaboration leading to even greater insight and continued progress against cancer.

About

Arin Karimian

Arin is the Corporate Content Lead at Leidos. He creates and curates content across a wide range of topics -- familiar territory for someone who's worked in banking, health care, media, and the non-profit space.

Follow on Twitter More Content by Arin Karimian
Previous Article
Meeting the Challenges of the Healthcare Revenue Cycle
Meeting the Challenges of the Healthcare Revenue Cycle

Successful and profitable healthcare institutions often have a firm grasp on every individual step, includi...

Next Article
Extending the Continuum of Care through the Path to Readiness
Extending the Continuum of Care through the Path to Readiness

Achieving a high level of care through an affiliate model requires a strong commitment and preparation to p...