NGFN-PLUS
Genetic epidemiology methods platform
| Coordinator: | Prof. Dr. Andreas Ziegler | |
| Institution: | Universität Lübeck | |
| Homepage: | www.imbs-luebeck.de |
The first aim of the genetic epidemiology methods platform was the biometrical support for all project partners within the consortium. In the process, we built on existing successful co-operations.
According to the latest development in genomics research regarding technology as well as statistical methods, our support was extended to different study types. Thus, in the first place, the project was successful in the identification of susceptibility genes in association studies of candidate genes. These candidates had been derived from previous linkage or association studies or in functional analyses. Secondly, in genome-wide association studies we were able to identify further genetic variants that are reliably associated with coronary artery disease and related phenotypes. To increase available sample sizes and thus improve the chance to detect variants even with smaller effects, we additionally performed meta-analyses of association studies within this project.
The second aim of this project was to facilitate statistical analyses of genome-wide association studies and the subsequent development of diagnostic and prognostic instruments. For this purpose we have developed a new implementation of machine learning algorithms that is now freely available in the software RandomJungle (www.randomjungle.de). The included methods have as yet not been part of the standard repertoire in analyzing genome-wide association studies and are sufficiently flexible to be applied to different types of data, such as common and rare genetic variants as well as clinical risk factors. They are specifically designed to correctly classify probands from different subgroups, such as with and without the disease; thus, they directly allow for the development of diagnostic and prognostic models. The implementation is time and CPU efficient, so that extremely large data sets can be analyzed. Additionally, the data sets can be scrutinized for interactions between variants and risk factors. Thus, they enable the transition from the mere detection of associated genetic regions towards the development of algorithms that can be utilized as tools for stratified medicine.
Further Coordinators:
According to the latest development in genomics research regarding technology as well as statistical methods, our support was extended to different study types. Thus, in the first place, the project was successful in the identification of susceptibility genes in association studies of candidate genes. These candidates had been derived from previous linkage or association studies or in functional analyses. Secondly, in genome-wide association studies we were able to identify further genetic variants that are reliably associated with coronary artery disease and related phenotypes. To increase available sample sizes and thus improve the chance to detect variants even with smaller effects, we additionally performed meta-analyses of association studies within this project.
The second aim of this project was to facilitate statistical analyses of genome-wide association studies and the subsequent development of diagnostic and prognostic instruments. For this purpose we have developed a new implementation of machine learning algorithms that is now freely available in the software RandomJungle (www.randomjungle.de). The included methods have as yet not been part of the standard repertoire in analyzing genome-wide association studies and are sufficiently flexible to be applied to different types of data, such as common and rare genetic variants as well as clinical risk factors. They are specifically designed to correctly classify probands from different subgroups, such as with and without the disease; thus, they directly allow for the development of diagnostic and prognostic models. The implementation is time and CPU efficient, so that extremely large data sets can be analyzed. Additionally, the data sets can be scrutinized for interactions between variants and risk factors. Thus, they enable the transition from the mere detection of associated genetic regions towards the development of algorithms that can be utilized as tools for stratified medicine.
KTT
MEDIA
CURRENT
NGFN-MEETING-2012
NGFN- MEETING
NGFN1&2
LINKS