We will use genotyping, whole exome sequencing (WES) and whole genome sequencing (WGS) data in UK Biobank as predictive variables in machine learning approaches to distinguish people with neurodegenerative diseases and healthy controls. The correctly classified genetic patterns built with the prioritized genomic variants in the models will be characterized in terms of interactions and association with comorbidities. We would like to show that machine learning methods are capable of prioritizing genomic variants not only based on the individual enrichment of each variant in different conditions, but also considering interactions between groups of variants. Thanks to the fact that UK Biobank is a valuable source of health data, we will retrieve the comorbidities that are related to the individuals with the neurodegenerative disease and the genomic interactions happening together. Methodologically speaking, machine learning methods can add an important layer of information to the disease related genomic variants obtained with other population genomic approaches such as genome-wide association studies (GWAS) and represent a new way to dissect genetic determinants of the disease. In this regard, the combination of these methodologies could advance the discovery of new genetic determinants predisposing to the disease and disentangle the biological mechanisms involved. Furthermore, the validation of these genetic determinants in other cohorts could improve the clinical characterization of patients with neurodegenerative diseases in the future. This work is part of my PhD project aimed to develop predictive models in neurodegenerative disease integrating various sources of data. The expected duration of the project is three years.
