001  Machine learning accurately determines the population prevalence of Alzheimer’s disease based on microbiome profile

The human microbiome is a complex and dynamic community of microbes, thought to have symbiotic benefit to its host. Influences of the gut microbiome on brain microglia have been identified as a potential mechanism contributing to neurodegenerative diseases such as Alzheimer’s disease (AD), motor neurone disease and Parkinson’s disease. We hypothesise that machine learning applied to the population level gut microbiome will predict the prevalence of AD. Analyses were performed in R, using two large, open-access microbiome datasets (n=1006 & n >2000) (Lahti et al., 2014; Pasolli et al., 2017). Countries in these datasets were grouped based on AD prevalence and the microbiome profiles compared. In countries with a high prevalence of AD, there is a significantly lower diversity of the gut microbiome (p<0.001). A PERMANOVA (p<0.05) revealed significant differences between several taxa in countries with high vs low prevalence of AD. Additionally, using machine learning, we were able to predict the preva- lence of AD within a country based on the microbiome profile (Mean AUC 0.935 & 0.870). We conclude that differences in the microbiome can predict the varying prevalence of Alzheimer’s disease between countries. Our results support a key role of the gut microbiome in neurodegeneration at a population level. a.basgaran@nhs.net