RT Journal Article SR Electronic T1 Does genetic risk help to predict amyloid burden in a non-demented population? A Bayesian approach JF bioRxiv FD Cold Spring Harbor Laboratory SP 174995 DO 10.1101/174995 A1 Nicola Voyle A1 Willemijn Jansen A1 Aoife Keohane A1 Hamel Patel A1 Amos Folarin A1 Stephen Newhouse A1 Caroline Johnston A1 Kuang Lin A1 Pieter Jelle Visser A1 Angela Hodges A1 Richard JB Dobson A1 Steven J Kiddle A1 for the Alzheimer’s Disease Neuroimaging Initiative A1 EDAR and DESCRIPA study groups YR 2017 UL http://biorxiv.org/content/early/2017/08/10/174995.abstract AB INTRODUCTION In this study we investigate the association between Aβ levels in cerebrospinal fluid (CSF) and genetic risk in a non-demented population. This paper presents the first analysis to use a Bayesian methodology in this area.METHODS Data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the EDAR* and DESCRIPA** studies was used in a Bayesian logistic regression analysis. We modeled CSF Aβ burden using age, diagnosis (healthy control or mild cognitive impairment), APOE and a polygenic risk score (PGRS) associated with Alzheimer’s Disease (AD). We compared models built using informative priors on age, diagnosis and APOE with non-informative priors on all variables.RESULTS The use of informative priors did not improve model performance in the majority of cases. Models using only age, diagnosis and APOE genotype showed the best predictive ability.DISCUSSION A previous study indicated that a PGRS of AD case/control status was associated with CSF Aβ burden in healthy controls. The current study suggests that this association does not lead to models that are more predictive of amyloid positivity than already known factors such as age and APOE.*‘Beta amyloid oligomers in the early diagnosis of AD and as marker for treatment response’**‘Development of screening guidelines and criteria for pre-dementia Alzheimers disease’