Comorbidity patterns in inflammatory skin diseases: A systems medicine approach using machine learning and omics technologies
Many patients with chronic inflammatory skin diseases, including atopic dermatitis and psoriasis, develop auto-immune mediated comorbidities affecting other barrier organs. However, the underlying pathophysiological and molecular mechanisms are largely unknown. The Junior Research Group will follow a systems medicine approach to improve our understanding of these cross-phenotype aetiologies, integrating multiple levels of omics data, clinical information and external biological knowledge in mathematical models that allow stratification of patients according to their comorbidity status. The main focus will be on the interpretability of the created models thereby enabling the in silico identification of key molecules, pathways and biological networks.
Machine-learning techniques will be employed to tackle the specific challenges of analysing high-dimensional omics data sets. Novel approaches will be developed to enrich the set of possible molecular risk factors for those that are relevant for classification, to incorporate information about structural and functional relationships between molecules into the model training step and to build an integrated model based on multiple levels of omics data. Extensive simulation studies will be performed to understand strengths and weaknesses of both novel and existing machine learning methods. Both, methods and the simulation framework, will be implemented as packages of the freely available and widely used statistical software R. Several patient collections with different types of omics data measured in whole blood or skin biopsies are available to the group. In collaboration with clinical partners, these collections will be extended further by specifically recruiting patients with selected comorbidities.
Degenhardt, F., Seifert, S., Szymczak, S., 2017. Evaluation of variable selection methods for random forests and omics data sets. Brief. Bioinformatics. https://doi.org/10.1093/bib/bbx124
Kachroo, P., Szymczak, S., Heinsen, F.-A., Forster, M., Bethune, J., Hemmrich-Stanisak, G., Baker, L., Schrappe, M., Stanulla, M., Franke, A., 2018. NGS-based methylation profiling differentiates TCF3-HLF and TCF3-PBX1 positive B-cell acute lymphoblastic leukemia. Epigenomics 10, 133–147. https://doi.org/10.2217/epi-2017-0080