Data-Management, -Integration, and -Modelling

Beule Lab

The core unit bioinformatics (CUBI), headed by Dieter Beule, comprises experts, who face the challenge of managing rapidly increasing amounts of complex molecular data and provide solutions and analyses that are usable for biomedical research and clinical application. The CUBI team focuses on method development and algorithms for efficient, reproducible, and reliable omics data processing and interpretation as well as new access data technologies. The omics data analysis comprises the areas of rare disease genetics, cancer genomics, functional genomics, single cell transcriptomics, as well as metabolomics and proteomics. CUBI already developed systems for omics data access and retrieval (SODAR) that allows handling of proteomics, metabolomics and genomics raw data with related pseudonymised metadata and in addition, processing results with related log-files.

As part of the MSTARS consortium, Dieter Beule and the CUBI team design new metadata models for the new data types derived from e.g. MALDI Imaging or CyTOF experiments. Subsequently, they establish data import pipelines for the new technologies into SODAR, followed by the validation of these pipelines to guarantee high data quality and consistency. The team of Dieter Beule similarly supports the data analyses of the different work packages. In addition, they perform data integration of omics data sets to retrieve signatures with biomedical relevance and compare the results with existing and validated methods for treatment decisions. In the long run, the usability of the generated information for the molecular tumor conference will be evaluated.

Blüthgen Lab

Nils Blüthgen’s lab uses mathematical modeling, bioinformatics and quantitative cell biology to investigate intracellular signaling and gene regulatory networks. Their prime research interest is how oncogenic signaling pathways and their downstream gene regulatory networks mediate their oncogenic potential, how drugs can modulate these networks and how resistance against targeted kinase inhibitors arises. The team analyzes and integrates quantitative or genome-wide experimental data using state-of-the-art bioinformatical methods and mathematic models.

In the MSTARS consortium, the Blüthgen lab develops semi-mechanistic, computer-aided models, which enable the analysis and integration of perturbation data of pre-clinical models to retrieve biomarkers or disease signatures that can be tested on clinical patient samples. First, the models are used to shed light on varying responses upon the inhibition of signaling pathways and to understand therapy resistances and associated mechanistic markers. This might result in refined treatment recommendations, which can be tested experimentally or in patient cohorts. Based on these data the models will be optimized iteratively. In addition, spatial data of interesting analytes are extracted from the MS- and cytometry-based imaging experiments and used to further refine the established models. In the end, a similar strategy will be applied to at least two other cancer entities.

Klauschen Lab

The work of the Klauschen lab is based on systems medicine approaches that integrate histo-morphological, genomic and proteomic analyses with the aim to improve the identification of clinically and therapeutically relevant molecular changes in tumors. The lab uses experimental methods and bioinformatics approaches, such as machine learning and simulations, to understand pathological characteristics of tumors and use this knowledge to predict better and targeted therapies for individual patients. To enable the integration of proteomic datasets the team works on methods to efficiently process and enrich these sensitive samples for proteomic mass spectrometric analyses. The combination of proteomic data with other omics data and their subsequent integration via network models gives insights in the functional relevance of mutation profiles and enables the simulation of effects of activating mutations and their inhibition by drug treatments.

In the context of the MSTARS consortium, the Klauschen lab uses machine learning and neuronal networks connected with layer-wise relevance propagation to predict clinically relevant characteristics of multi-omics and histological data. The team develops new machine learning approaches that allow the analysis of clinical cohorts and experimental models and help to answer the questions of whether metabolomic can predict proteomic data and vice versa, as well as whether methylation patterns can be predicted from proteomic data. In addition, they will adapt the machine learning approaches to cope with heterogeneous input data from imaging-based, histological and multi-omics analyses and will address the more global question on how imaging-based and histological data can complete multi-omics data to predict clinical parameters and therapy responses.