Computer-Aided Decoding of Brain-Immune Interactions in Gulf War Illness (GWI): A Joint Embedding on Brain Connectomic and Immunogenomic Markers
Abstract
At least 25%-30% of the nearly 700,000 U.S. Veterans who served in the 1991 Gulf War continue to suffer from a complex, multisystem disorder called Gulf War Illness (GWI). The exact cause of GWI remains unknown, and efforts directed towards developing treatments for GWI have been hampered by the lack of objective tests and biomarkers for the illness. Data suggest that there is a strong brain-immune component to the disorder and that biological measures taken from brain imaging and immune-genetic measures (blood, cerebrospinal fluid) reflect different but potentially connected aspects of the illness. These measures have, mainly, only been studied in isolation and the predictive risk of the illness at the individual Veteran level has not been high enough to be useful. Combining information across different biological markers into a joint distribution or combined risk factor offers the potential for better describing the cause of GWI over the one based on a single marker or simply adding together of different markers. However, this requires special tools to do this type of highly complex computer computations. Persisting symptoms of GWI include a heightened or chronic inflammation in the brain. This is unlike a typical inflammatory illness that resolves with a slow but steady recovery of the immune system over time. At the same time, alterations have been described in the brains of Veterans with GWI using various brain imaging techniques. This converging information suggests that a better understanding of brain-immune interactions may provide a key to unlocking the biological origin of GWI and is the focus of research for the large, Department of Defense-funded multi-site Boston GWI Consortium (GWIC) studies (GW120037). Identifying the source of immune activation and brain imaging biomarkers has been recommended by government advisory groups as important research targets for GWI. However, traditional analytic models to define biomarkers for GWI have failed to draw a complete figure of the complex brain-immune interactions underlying the illness. The central goal of this project is to build a comprehensive computerized classification model of GWI that includes all of these markers into one model. We will utilize a machine learning framework, which is derived from artificial intelligence technology, to incorporate different biomarkers (blood tests, cerebrospinal fluid, brain imaging) for further investigation of the complex interactions that represent GWI etiology. This approach offers major advantages over approaches based on a single marker or a simple adding together of different markers within a single biological domain. Machine learning technology has already been applied to aid our everyday life. These methods are routinely used to perform functions ranging from predicting the track of storms to identifying fraudulent money transactions. Machine learning methodologies have been essential for identifying meaningful patterns within the massive amount of data we collect. In this project, we will apply machine learning to incorporate multiple biomarker data in order to generate probability scores at the individual Veteran level to determine who has a potential risk for symptoms of GWI. This will be done by using data from a known cohort of GW Veterans from the Boston GWIC studies. The goal of this study is to take the various promising biological markers of GWI from the GWIC studies where the overall hypothesis is that GWI is associated with altered brain-immune interactions and cross-talk pathways. Biomarkers of GWI are currently being collected from 250 GW Veterans (150 cases, 100 controls) for immune, genetic, and neuroimaging markers in the GWIC. Cases and controls are determined by both Kansas and Centers for Disease Control and Prevention (CDC) GWI criteria. This study will bring talented bioengineers to the GWIC team to combine the enormous amount of GWIC biomarker data into a cutting-edge 3D mac
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 07, 2017
- Source ID
- W81XWH1710440
Entities
People
- Bang-Bon Koo
Organizations
- Boston University Medical Campus
- United States Army