Machine Learning Tools to Identify and Associate Genetic Variants in Patients With Phenotypic Traits of Brain Injury Associated Fatigue and Altered Cognition (BIAFAC)

Annually 1.5 million children and adults experience trauma to the head and brain that results in a TBI. Our research suggests that in a subset of patients, TBI induces pituitary dysfunction and abnormal growth hormone (GH) secretion. The clinical syndrome associated with abnormal GH secretion is characterized by profound fatigue and cognitive dysfunction related to executive function, short-term memory, and processing speed index. Fatigue in these patients is profound and debilitating leaving them unable to maintain their usual activity levels. We have termed this syndrome Brain Injury Associated Fatigue and Altered Cognition (BIAFAC).

Our recent work has shown that cognitive and physical dysfunction are significantly improved with recombinant human growth hormone replacement in patients with BIAFAC. Improvements in fatigue often precede (~3 months) improvements in cognition (~4-5 months) following rhGH treatment. Although rhGH replacement relieves BIAFAC symptoms, it does not cure the underlying cause, as symptoms reoccur with rhGH withdrawal.

Although the mechanisms causing BIAFAC have not been determined, our previous research demonstrated that a year of GH treatment resulted in symptom relief which was associated with changes in brain morphometry and connectivity. These associated brain changes include increased frontal cortical thickness and gray matter volume as well as resting state connectivity changes in regions associated with somatosensory networks

The next step to understanding BIAFAC is to develop a biomarker that identifies individuals that are susceptible to developing this syndrome. The University of Michigan maintains a searchable DataDirect database of over 4 million individual patient medical records linked via the Michigan Genomics Initiative (MGI) to genomic data collected from over 70,000 patients. By collaborating with the University of Michigan, we have a unique opportunity to combine their extensive genomic database with the more than 100 UTMB patients we are currently treating for BIAFAC to search for common genetic markers associated with BIAFAC. In order to identify patients in the UM genomic database with BIAFAC, we will develop a risk stratified machine-learning algorithm based on BIAFAC symptoms. Initial use of the algorithm will begin with approximately 9,000 patients in the UM database that have already been identified with a diagnosis code of fatigue and malaise. Once these patients are identified, a select cohort will be contacted to confirm the accuracy of the algorithm in identifying BIAFAC patients. Once we complete the genotyping of UTMB patients with BIAFAC and have identified the patients with BIAFAC in the UM genomic database, a genome-wide association study (GWAS) will be executed to look for common genetic markers

Aims: