Abstract:
|
In emerging research fields, such as imaging genetics research and multi-level omics research, researchers often deal with not only a large number of genetic variables but also high-dimensional multivariate outcome data. While conventional methods can be applied to analyze one outcome or one genetic variant at a time, there is a pressing need for new analytical methods to simultaneously model high-dimensional genetic and outcome data. Furthermore, the underlying relationship between genetic variants and outcomes is usually complicated and unknown, which adds another layer of difficulty to data analysis. To address these challenges, we propose a functional neural network (FNN) method for high-dimensional genetic data analysis with single or multiple outcomes. FNN uses basis functions to model high-dimensional genetic and outcome data, and further builds multi-layer functional neural network to capture the complex relationship between genetic variants and outcomes. Through simulations, we show FNN outperform conventional methods, such as functional linear models.
|