The analysis of genome-wide association studies is the standard tool to identify a genetic causes of chronic diseases. In this project, a network-based kernel was developed that converts the genomic information of two individuals into a quantitative value reflecting their genetic similarity. The kernel integrates the network structure of biological pathways as prior knowledge into the analysis. Here, a pathway is defined as a network of interacting genes responsible for achieving a specific cell function or regulation. The benefit is the potential interpretation of the disease association in a biological context and reduction of a number of statistical problems. The approach is exemplified to genome-wide association case-control data on lung cancer and rheumatoid arthritis. Some promising new pathways associated with these diseases are identified, which may improve our current understanding of the genetic mechanisms.
- descriptive network analysis
- network-based kernel construction
- logistic kernel machine test that assumes a semi-parametric logistic regression model and tests the genetic effects with a score-type statistic.
Open-source software as R package kangar00 (Kernel Approaches for Non-linear Genetic Association Regression) is available on request. The package allows parallel computing of kernel matrices using the power of graphics processing unit (GPU).
- S. Freytag , J. Manitz, M. Schlather, T. Kneib, C. I. Amos, A. Risch, J. Chang-Claude, J. Heinrich, and H. Bickeböller (2013): A Network-Based Kernel Machine Test for the Identification of Risk Pathways in Genome-Wide Association Studies. Human Heredity, 76(2), pp. 64-75. Shared first co-auhtorship.
- J. Manitz, S. Friedrichs, B. Hofner, P. Burger, contributions by S. Freytag, N.-T. Ha, M. Schlather, and H. Bickeböller (2015). kangar00: An R package for Kernel Approaches for Nonlinear Genetic Association Regression. Available on request.
- Invited talk at Workshop “Statistical Network Science and its Applications” Cambridge, UK, in August 2016.