Identification of causal risk factors for complex inflammatory diseases by convolutional neural network modeling of single-cell omics data

Abstract

For the last decade, many variants on chromosomes related to various diseases have been discovered along with the development of Genome-wide association studies (GWASs). However, it is challenging to specify variants that directly affect the disease due to linkage disequilibrium (LD). Also, which cell type a causal variant functions as a risk factor remains a task to be solved. Therefore, I developed a convolutional neural network (CNN) model that indicates variants and cell types directly related to diseases using GWAS summary statistics and information about open chromatin regions of several cell types from single-cell OMICS data. The model is expected to find novel causal risk factors which previous fine-mapping methods could not detect. Also, some genes will be presented as drug targets through downstream analysis to connect the causal risk factors to associated genes.