Genetics of deep-learning-derived neuroimaging endophenotypes for Alzheimer's Disease

Abstract There is still a lack of knowledge on the key genetic factors associated with post acute syndrome for COVID-19 patients, especially those related to neurological complications. In this study, we will utilize both the genetic and clinical data available through the All of Us researchers platform to study the genetic association with COVID-19 complications. In order to more accurately phenotype the patients based on their clinical trajectory mostly recorded in their electronic health records, we will utilize a pretrained deep learning model trained on more than four million patients from the N3C cohort. The pretrained model will be further fine-tuned on the All of US data, and will be used to phenotype the patients with genetic data. Further GWAS study will be performed to correlate between the deep learning based phenotype and the genetic information. Successful completion of this project will bring new insights to guide COVID-19 patients treatment to better prevent or manage further complications.