Hack-a-thon Reveals Unexpected Collaboration with Airline Industry
This article was originally published on Wednesday, October 27 as a guest blog entry in Cornerstone, a publication of the Children’s Hospital of Philadelphia’s Research Institute.
Editor’s Note: Since its inception, the Children’s Brain Tumor Network (CBTN) has been committed to global inclusion, open science, and a pioneering spirit in its mission to deliver better therapies for children with brain and central nervous system tumors. These commitments are shared across CBTN’s global network and coordinated by the network’s operations center at Children’s Hospital of Philadelphia’s Center for Data-Driven Discovery in Biomedicine (D3b). In this guest blog, Jonathan Waller, a senior communications specialist with the D3b, shares how they furthered their mission to advance research at the third annual Hack4Rare rare disease hack-a-thon hosted by the Children’s Tumor Foundation and MIT Hacking Medicine.
Teamwork between the Center for Data-Driven Discovery in Biomedicine (D3b) bioinformaticians, CHOP’s Cancer Center, and American Airlines’ machine learning computer scientists earned a major award at a rare disease hack-a-thon in which they devoted their unique expertise in data science for the airline industry to solve a complex question in pediatric cancer research.
The American Airlines team, Team American, had no background in bioinformatics, oncology research, or biology. But what they did have was powerful artificial intelligence developed to organize and analyze data for the airline — and a spirit of volunteerism and desire to help children. Through the competition, they made a new discovery that could have a significant positive impact on the future of treatment development for children with neurofibromatosis (NF).
At the third annual Hack4Rare event, in which the D3b and the Children’s Brain Tumor Network (CBTN) served as data contributors and project mentors, more than 300 data and computer scientists across 32 teams competed in the development of solutions with real-world applications for NF, PTEN hamartoma tumor syndrome, RASopathies, and desmoid tumors. The goal for all participants was to drive scientific and medical innovation and improve the lives of patients living with rare diseases.
D3b data experts, in partnership with CHOP research scientist, Thomas De Raedt, PhD, had the privilege to mentor a team of machine learning software engineers from American Airlines in their analysis of NF tumor data. While this team of experts was not an obvious match for this sort of competition, they worked together over the course of five weeks at the CBTN operations center at Children’s Hospital of Philadelphia to examine the raw data available. Team American chose to use a machine learning program they had developed and applied it to open-source data within the CBTN’s Pediatric Brain Tumor Atlas (PBTA). They consulted with CHOP-based bioinformaticians leading CBTN’s OpenPBTA initiative, posed questions about the data, and engineered ways to apply their artificial intelligence tools to yield usable results.
Team American won third prize for their project, “Graph neural-network based gene clustering for neurofibromatosis patients,” which applied a new algorithm to challenge current methodologies to analyze genomic data. Instead of summarizing data starting from the gene sample level, they started from the transcript level data. With their machine learning program, the team found they were able to develop a new framework to subtype pediatric NF.
“We were not expecting this outcome,” said team leader Deepak Agrawal. “A team member had a personal connection with pediatric cancer and inspired us to participate. We wanted to do whatever we could do to advance research in this field.”
In the field of precision medicine, this is an invaluable finding. Having the ability to more specifically diagnose a patient’s cancer can enable doctors to develop specialized treatment plans that are best suited to the patient. The team’s discovery process was iterative, and it provides a strong case study for how data scientists across industries can collaborate to solve complex problems.
More Partners Needed
The success of Team American within the Hack4Rare event highlights the power of innovative thinking and multidisciplinary collaboration. Under a siloed model, a single researcher can sometimes require five years or more to bring together enough data or specimens to start making notable discoveries. However, since 2011 over 150 data projects and 100 biospecimen projects have been launched with support from CBTN and its biospecimen and data resources.
“Many thanks to CHOP, who met with us every week during the competition to help us refine our strategy for solving this problem and informed us about the research that had already been done so we wouldn’t be reinventing the wheel,” Agrawal said. “The OpenPBTA initiative is doing an amazing job of making the genomic patient data open source. There is so much community support.”
Team American’s participation in this hack-a-thon event was something of an experiment. And like any successful scientific experiment, the results can be repeated. There is so much data left to work through, and more is on the way. By next year, the more than 4,200 patient tumor specimens within the CBTN biobank will be sequenced. That will result in a hitherto unheard-of cache of raw omic data related to pediatric brain and spinal cord tumors, available at no cost to researchers and clinicians across the globe.
Within this huge trove of information lay both challenges and opportunities. The challenge lies in getting these data into enough hands to organize and analyze them, so they can be useful to the research community. The opportunity is for collaboration — with innovators in any industry, at any career level, anywhere in the world — to make new discoveries that can change the trajectory of pediatric neuro-oncology research and in treatments for children. By harnessing the brain power of our global community and leveraging new technologies, we can uncover helpful knowledge faster than ever before.