Identification of CSF-based molecules predictive of changes related to early cognitive impairment

Bodily fluids such as cerebrospinal fluid, which is in direct contact with the brain are rich sources of biochemical measurements such as proteins and metabolites. Abnormal levels of these measurements can serve as biomarkers for certain conditions, such as Alzheimer's Disease (AD). Previously, CSF-based biomarkers for diagnosing AD have been developed, however comparatively little research has been conducted into early detection of AD using these measurements. In this project, we utilize publicly available data from the Alzheimer's Disease Neuroimaging Initiative to 1) determine if there is a relationship between CSF biomarkers and psychiatric evaluations or anatomical changes in brain regions and 2) to determine if there are differences in these CSF-based biomarkers in cognitively normal (CN) or early mild cognitively impaired (EMCI) patients. We analyzed spearman correlation matrices between the CSF data and survey/brain region and they indicated that there was no linear association between the data sets. As such, we developed machine learning models that predict psychiatric survey and MRI brain region outcomes using CSF metabolite concentrations. In order to develop a model with the highest accuracy we have tested various regression and feature selection models to identify the optimal predictor brain region and survey data. Further, separate analyses are being conducted for CN and EMCI patients to determine changes in metabolite concentrations associated with the earliest symptoms of cognitive impairment.