Causal network inference with application to breast cancer

复杂疾病通常涉及DNA序列、转录和表观遗传过程的变化，如

Complex diseases often involve changes in DNA sequence, transcription, and epigenetic processes such as methylation. These changes lead to a wide range of symptoms or multiple subtypes of the same disease. In order to develop more effective treatments for different disease subtypes, we need to better understand the genes and processes (i.e., transcription and methylation) that drive these differences. Unfortunately, identification of genes and processes that underlie a disease is often compromised by inference based on correlation, not causation. Our long-term goal is to develop computational methods to infer gene regulatory networks that are causal for multiple clinical phenotypes using genomic and clinical data of complex diseases. In this project, we will develop and test new statistical approaches to identify regulatory networks involving both transcription and methylation that are potentially causal for disease subtypes. Our strategy is to use the principle of Mendelian randomization. This assumes that the alleles of a genetic variant are randomly assigned to individuals in a population, analogous to a natural perturbation experiment. Whereas most existing methods for studying interactions among genes look at correlation (or association), this principle allows us to separate correlation due to causation from correlation not due to causation. We will develop our approaches via three specific aims and will use breast cancer as the disease model: (1) Develop a causal network model using genotypes, expression and methylation of single genes. (2) Develop a causal network model to identify individual genes whose transcription or methylation is causal for multiple clinical phenotypes. (3) Develop a causal network model to identify combinations of genes whose transcription or methylation are causal for multiple clinical phenotypes. The models and algorithms developed in this proposal will allow us to make causal statements about the two processes at multiple genes and account for confounding variables, neither of which has been examined before in similar studies. These models will identify key genes for specific breast cancer subtypes and the roles for transcription and methylation when many genes are involved, leading to better diagnoses and development of novel drug targets.