Speeding Up Cardiac Disease Research with Big Data

How can adopting big data strategies from the business world help the medical field advance research?

The business world has used big data to drive decisions for years. Now, doctors are bringing this approach from the board room to the lab bench, using data to inform research questions. Investigators in Children’s Hospital Colorado’s Heart Institute are using big data and multiomics profiling as a launchpad to uncover new information and hypotheses about pediatric heart disease. This technique can point researchers toward new areas of study and accelerate the research process to impact patients faster. As Ben Frank, MD, puts it, “Sometimes you don’t know what you don’t know until you find it.”

Inductive research approach for cardiac research

Jesse Davidson, MD, and his team started exploring this inductive research approach at Children’s Colorado in 2018 with a metabolic profiling study of infants under 4 months old who underwent cardiothoracic surgery. At the time, Dr. Davidson says, this was one of the first major pediatric medical studies to use the data-driven approach. The team’s work showed significant dysregulation in many of the metabolic pathways, pointing them to areas of further research.

The big data process looks slightly different each time, but the research typically starts with blood tests. If Drs. Davidson and Frank are using a metabolomics approach, that means measuring anywhere from 200 to 250 biomarkers. If they are employing a proteomics approach, that involves measuring closer to 1,500 markers in each patient. Then, the researchers plug that data into a machine-learning algorithm that sorts through what might be significant depending on how the researchers are dividing up the population they are studying, such as kids with heart disease compared to kids who don’t have heart disease. After that, they are able to ask questions and test hypotheses.

“Zooming out to see the truth can help you see things you weren’t expecting,” Dr. Davidson says. “Science has traditionally been a hypothesis-driven product. You make a hypothesis about a next experience, test that hypothesis, come up with more data, make your hypothesis, test that, etc. It has blinders on to just about everything else that’s going on. It is extraordinarily challenging to work down a very linear path and expect that you’ll know a lot about the rest of the system.”

While Dr. Davidson says this more traditional, hypothesis-driven approach is still extremely important in the research process, using data to drive the hypotheses can account for more complexity.

As Dr. Davidson, Dr. Frank and their colleagues started seeing clinical research yield more complex and intricate data, they realized this type of big data approach would be ideal, especially for their work with patients with single ventricle heart disease. The team treats children with this condition regularly, so there’s a trove of clinical data to analyze.

“There’s a lot of room to do better, and there’s a lot of room to help their lives, to help their survival and to help their day-to-day experience,” Dr. Frank says. “We are at a unique and powerful moment to be at the forefront of trying to figure out how to leverage this new data, and how to use it to help kids.”

Dr. Davidson agrees: “Cardiology and cardiac surgery are ripe for this reverse strategy. If you start with an inductive approach, are there ways to find smarter ways forward and also to speed up the discovery process?”

The orchestra behind big data analytics

Big data analytics would not be possible without a team of experts working together across the University of Colorado (CU) Anschutz Medical Campus. For example, as part of this work, it was necessary to engage with statisticians, and CU is home to the Center for Innovative Design and Analysis — an invaluable tool for researchers partnering with biostatisticians and data scientists on complex data sets. Drs. Davidson and Frank also work closely with a team of research coordinators, research nurses and regulatory support teams.

“Sometimes I think Dr. Davidson and I are conductors of a big orchestra. Without all the instruments, though, you wouldn’t get any music,” Dr. Frank says. “That kind of collaborative, team-based approach has been essential to our success.”

This collaborative style of research allows the team to follow the path laid out by the data. Dr. Davidson and his team are taking this work one step further, by not just generating new hypotheses after analyzing the data, but also dedicating time and studies to test those new, specific hypotheses to see what they learn.

“I think that’s where Dr. Davidson has really led,” Dr. Frank says. “If you look at the machine-learning literature, there’s a lot out there about using these strategies to do hypothesis-generating studies in different populations, but there are fewer people who are pushing to take the next step: hypothesis-testing validation of their findings. That next step is the crucial piece to move our work closer to improving outcomes for kids at the bedside.”

Metabolomics and phenotypic approach

Dr. Davidson’s 2018 Journal of the American Heart Association study laid the groundwork for this type of data analysis research to blossom in the Heart Institute. Through that research, which took a metabolomic approach, Dr. Davidson and his team found a profound shift in the metabolic fingerprint of infants undergoing cardiothoracic surgery with cardiopulmonary bypass. They also noted a global deficiency in amino acid levels, which is a finding that can lead to quick, direct changes to the care patients receive and an area the team will focus on next.

The team also used metabolomic profiling in a 2022 paper in the American Journal of Renal Physiology where they explored acute kidney injury, a common cause of morbidity after congenital heart disease surgery. Using animal models, the researchers were able to identify novel evidence of dysregulated tryptophan catabolism, among additional findings, opening the door to explore these pathways for diagnostic and therapeutic targets.

In a 2023 Journal of the American College of Cardiology: Advances paper, Dr. Frank analyzed how kids with single ventricle heart disease are different than other populations, such as those without heart disease, and how doctors can be more personalized in their approach to treating these children. The findings suggested an association between increased pre- and post-operative circulating methionine and tryptophan metabolite levels, as well as difficulty recovering from heart surgery. These three examples are just a handful of many research projects the Heart Institute team has conducted with this big data approach.

“I think taking the big data approach allows you to build this really deep phenotype of who these kids are at these different key, crucial moments of leverage in their life,” Dr. Frank says. “This type of approach really allows you to blend the acute and the chronic. It really allows you to ask questions about both how we can help you get through this moment of crisis in your life, and also when you are OK, how we can help you maintain that functional status? How can we make you feel even better that you thought was possible?”