Research Projects

The central goal in this project area is the modelling of health indicators with modern regression methods. The influence of high-dimensional omics data on the relationship between toxicological exposure and health is investigated. The models considered comprise classical linear and logistic regression as well as modern procedures such as logic regression, regularised procedures and generalised additive models. For selecting the most important influential variables and interactions, methods like component-wise boosting will be generalised, genetic risk scores will be formulated, and sampling techniques will be adapted.

• R1 High-dimensional regression for screening of important genetic and environmental factors (Ickstadt, Schikowski)
• R2 Regression with nonlinear modelling of metric environmental and toxicological influence factors (Groll, Schikowski)
• R3 Statistical assessment of gene-exposure and gene-gene interactions (Schwender, Schikowski)

A central topic in this project area is the estimation of the minimum effective dose of a compound, using high-dimensional omics data as a surrogate for effects on health. Statistical learning methods are used to select relevant genetic markers. In addition, multi-level statistical modelling approaches will be developed for the simultaneous consideration of different doses and exposure times, and methods for deriving optimal experimental designs for such models.

• P1 Determination of the minimum effective dose from high-dimensional expression data with statistical learning methods (Rahnenführer, Hengstler)
• P2 Model selection for exposure time-dose curves with mixture models (Rahnenführer, Hengstler)
• P3 Two-dimensional spline regression to model incubation time-dose curves (Ickstadt, Krutmann)
• P4 Model-based interaction analysis of secretome and proteome data from intrinsically and extrinsically aged cells (Schwender, Krutmann)
• P5 Experimental design for exposure time-dose curves (Schorning, Hengstler)
• P6 Design and analysis of concentration-exposure curves with common parameters (Schorning, Cadenas)

• P7 Statistical methods for comparing gene expression curves of several genes or treatments  (Möllenhoff, Hengstler)

The aim of this project area is to improve toxicological risk prediction and the differentiation of toxicological effects through appropriate integration of data, models, and results. For data integration, various molecular data sources will be combined and additional biological knowledge will be integrated into the modelling. Integrative modelling approaches will build upon models comprising those developed in the research areas regression and prediction.

• I1 Weighted risk scores for the assessment of cumulative risks (Ickstadt, Krutmann)
• I2 Prediction of phenotypical responses - from model systems to human diseases (Ickstadt, Cadenas)
• I3 Enrichment analysis for multiple data sources and complex statistical designs (Rahnenführer, Cadenas)
• I4 Integrative analysis of longitudinal omics data (Schwender, Hengstler)
• I5 Integration of different omics data with regression methods (Groll, Hengstler)