Count Phenotype-gene Association Detection Based On Zero-inflated Generalized Poisson Mode

Zero-inflated count phenotype data are increasingly common in gene genetic association studies,and mapping related genes by means of the data is an important task in current association studies.In developing genetic genetics,it is critical to use gene association detection to screen for possible disease-causing genes.Most common diseases are caused by multiple genetic and environmental factors that have a large impact on disease susceptibility.In monogenic diseases,the presence or absence of a disease allele can be used to predict the presence or absence of the disease.Diseases caused by a combination of multiple genes as well as environmental factors and monogenic diseases are inherited in different ways.Focusing on zero-inflated count phenotype data,this paper uses an EM adaptive LASSO method to screen possible causative genes and genetic factors based on a zero-inflated generalized Poisson model.The model and method combine the EM algorithm and the adaptive LASSO method,which has an advantage of allowing simultaneous parameter estimation and variable selection.And it uses the Bayesian information criterion to determine the optimal parameters at the end.In certain conditions,the EM adaptive LASSO method has an Oracle property,which means that the estimated results obtained after using the method are the same as the final filtered variables and the results we expected in advance.Therefore,it is possible to correctly screen for potential causative genes and genetic factors among many genes and genetic variables by using the a zero-inflated generalized Poisson model.In this paper,we can analyze multiple variables simultaneously in a zero-inflated generalized Poisson model as well as using the EM adaptive LASSO approach,which changes the dilemma of analyzing only one variable at a time.In addition,we evaluated the limited-sample performance of the model and method through extensive simulation studies and used it to analyze actual mouse cholesterol stone data.We have implemented the method in the software program R.