Data scientist helps researchers pinpoint the microbiome's role in disease

August 30, 2018
Microorganisms live within and upon the human body. How they affect health is just beginning to be understood.
Microorganisms live within and upon the human body. How they affect health is just beginning to be understood.

Deep within the human body, the microbiome is the latest frontier in biomedical research. Trillions of bacteria, fungi, viruses and parasites live within and upon us, and these microbes differ from person to person, as well as from body part to body part within the same person.

Researchers are just beginning to understand how the microbiome can influence the health of its host, and, as exploration begins, researchers need a way to assess the microbiome’s effect. 

Alexander Alekseyenko, Ph.D., associate professor of biomedical informatics in the Department of Public Health Sciences, has worked on a number of projects analyzing microbiome data. His experience showed him there are gaps in the existing methodologies. 

Analyzing the effect of the microbiome is difficult because there are so many organisms involved, not to mention multiple variables coming from the human host. Each person’s gene expression, genetic composition and medical history is different.

Alekseyenko is working on improving methods of microbiome analysis as well as figuring out a way to chart the relationship between the microbes and the multiple variables coming from the host in an effort to determine how they relate to human diseases. 

The National Institutes of Health recently awarded him a $1.3 million grant, his first NIH research project grant. 

One of the goals of the project is to develop interactive tools that would allow a biomedical researcher not well-versed in statistics to conduct such analysis correctly on his or her own. He already has a couple of prototypes of interactive tools for researchers to conduct this type of analysis. 

To develop the methodology, Alekseyenko is partnering with five research projects, each of which is collecting microbiome data: 

  • žAt MUSC, Kristin Wallace, Ph.D., is studying racial disparities in colorectal cancer.

  • Also at MUSC, Charlie Strange, M.D., is the principal investigator for the Alpha-1 Foundation Research Registry, the largest registry of individuals with alpha-1 antitrypsin deficiency in the world, and is an investigator with a National Institutes of Health, National Heart, Lung and Blood Institute study. 

  • žAt New York University, two research projects are studying pancreatic cancer and lupus.

  • žAt Virginia Commonwealth University, researchers are studying pregnancy and preterm birth. 

Alekseyenko, whose initial work was supported by a College of Medicine Enhancement of Team Science Award, said the statistical framework for each research project is the same, but how it’s applied and interpreted would be different. 

Alekseyenko said the principle difference of his approach is that many bioinformatics analyses in the past have started at the bottom, comparing variables one by one and then adding on. He’s starting at the top, looking to see if there’s a difference when all variables are considered simultaneously, then stripping away nonessential variables to see if the differences remain.

Conceptually, he starts with pretty basic questions. For example, is there a difference between sick people and healthy people? He then looks to see whether that question can be answered in the aggregate, rather than trying to compare each individual microbe in the data set. This could be important in pilot studies, he said, because they typically have few subjects and therefore limited statistical power to detect changes at a univariate level. Similarly, if the individual microbes each have a small effect, each effect might not be noticeable unless they’re aggregated and analyzed with one of the methods his team is developing, Alekseyenko said. 

The role of the microbiome is often best understood as a mediator, acting as the go-between for some intervention, whether it’s medical treatment or diet, and how the body responds, which could show up as a change in health or weight, for example. There are many methods for modeling the effect of a mediator if the mediator is a single variable, he said. 

But when the mediator is highly multivariate and complex, such as the microbiome, it gets trickier. Can researchers answer whether an intervention’s effect on the patient has somehow been amplified or reduced by those trillions of bacteria that have colonized the site? Alekseyenko is finishing up the first paper now that addresses this question statistically.