Explaining the Genetic Causality for Complex Phenotype via Deep Association Kernel Learning

• Main Authors: Feng Bao, Yue Deng, Mulong Du, Zhiquan Ren, Sen Wan, Kenny Ye Liang, Shaohua Liu, Bo Wang, Junyi Xin, Feng Chen, David C. Christiani, Meilin Wang, Qionghai Dai
• Format: Article
• Language: English
• Published: Elsevier 2020-09-01
• Series: Patterns
• Subjects: kernel learning; association analysis; deep learning; genome-wide association studies; disease causality
• Online Access: http://www.sciencedirect.com/science/article/pii/S2666389920300684

Summary: The genetic effect explains the causality from genetic mutations to the development of complex diseases. Existing genome-wide association study (GWAS) approaches are always built under a linear assumption, restricting their generalization in dissecting complicated causality such as the recessive genetic effect. Therefore, a sophisticated and general GWAS model that can work with different types of genetic effects is highly desired. Here, we introduce a deep association kernel learning (DAK) model to enable automatic causal genotype encoding for GWAS at pathway level. DAK can detect both common and rare variants with complicated genetic effects where existing approaches fail. When applied to four real-world GWAS datasets including cancers and schizophrenia, our DAK discovered potential casual pathways, including the association between dilated cardiomyopathy pathway and schizophrenia. The Bigger Picture: Genetic mutations cause complex diseases in many different ways. Comprehensively identifying the genetic causality can lead to valuable insights into the development and treatment of diseases. However, existing genome-wide association study (GWAS) approaches are always built under linear assumption and simple disease models, restricting their generalization in discovering the complicated causality. DAK (deep association kernel learning) is a GWAS method that is constructed in a deep-learning framework and can simultaneously identify multiple types of genetic causalities without any modifications to the model. For biological contributions, the proposed approach enables the understanding of non-linear, complex genetic causalities and improves functional studies of the disease; for computational contributions, our method unifies kernel learning and association analysis in a joint explainable deep-learning framework.