Junior Research Group Computational Genome Biology
Dr. Bernd Fischer
Our group uses computational statistics and machine learning methods to understand and predict genotype to phenotype relationships. Our aim is to derive systems-level maps of interactions of genes and proteins in order to model and predict the functional consequences of sequence variations in the human population and in cancer. With the tremendous increase in the amount of biological data recorded with current high-throughput techniques, it gets more and more challenging to mine and interpret the available data. Our contribution is the development of computational statistics for high-dimensional data including feature selection tools, differential testing and network modeling. A research focus is in the modeling of genetic and physical interactions. Genetic interactions describe how genetic perturbations influence each other, e.g. with respect to their cellular phenotypes. We developed method for quality control, normalization and statistical analysis of image-based genetic interaction screens. To better understand the molecular mechanisms it is as well important to study physical interactions between molecules. We developed analysis methods for protein quantification by mass spectrometry data. With this and downstream bioinformatics workflows we contributed to get a better understanding of RNA-binding proteins. The originated software packages are made public available as open source mainly as contribution to the Bioconductor project.
We will develop computational and statistical methods required for the analysis of emerging, cutting-edge data, especially from high-throughput functional and genetic assays and advanced applications of spectrometry. Specifically, the computational research in proteomics will focus on integrating individual genome sequence into the analysis of proteomic mass spectrometry data. Furthermore, we will develop the computational methods needed to understand how gene products and their mutations are interacting with each other physically and functionally, and how genetic and environmental variables combinatorially result in phenotypes including disease. We will develop computational methods with the aim to predict the phenotypic outcome of drug treatments and combinations of drugs under a known genetic background. For this aim we will combine observational population data with high- throughput perturbation screens. The computational methods will be instrumental for cancer proteomics in basic and biomarker research and for the construction of predictive, translational platforms for the next generation of individualized, genome-informed drug therapies.
Horn, T., Sandmann, T., Fischer, B., Axelsson, E., Huber, W., & Boutros, M. (2011). Mapping of signaling networks through synthetic genetic interaction analysis by RNAi. Nature Methods, 8, 341-6
Castello, A., Fischer, B., Eichelbaum, K., Horos, R., Beckmann, B.M., Strein, C., Davey, N.E., Humphreys, D.T., Preiss, T., Steinmetz, L.M., Krijgsveld, J. & Hentze, M.W. (2012). Insights into RNA Biology from an Atlas of Mammalian mRNA-Binding Proteins. Cell, 149, 1393-1406
Fischer, B., Grossmann, j., Roth, V., Gruissem, W., Baginsky, S. & Buhmann, J.M. (2006). Semi-supervised LC/MS alignment for differential proteomics. Bioinformatics, 22, e132-140
Fischer, B., Roth, V., Roos, F., Grossmann, J., Baginsky, S., Widmayer, P., Gruissem, W. & Buhmann, J.M. (2005). NovoHMM: a hidden Markov model for de novo peptide sequencing. Analytical Chemistry, 77, 7265-7273