Epigenomics of non-small cell lung cancer and colon cancer
Non-Small Cell Lung Cancer (NSCLC) is primarily treated with radiation, surgery and platinum-based drugs like carboplatin and the major challenge is intrinsic or acquired resistance to chemotherapy. Molecular markers predicting the outcome for the patients are urgently needed. In the BMBF-funded project EPITREAT we employed patient-derived xenografts (PDX) to detect predictive methylation biomarkers for platin-based therapies. Using MeDIP-seq we generated genome-wide DNA methylation profiles of PDXs and identified a set of candidate regions with methylation correlated to carboplatin response and corresponding inverse gene expression pattern even before therapy (Fig. 1). This analysis led to the identification of a promoter CpG island methylation of LDL Receptor Related Protein 12 (LRP12) associated with increased resistance to carboplatin. Validation in an independent patient cohort confirmed that LRP12 methylation status is predictive for therapeutic response of NSCLC patients to platin therapy with a sensitivity of 80% and specificity of 84% (p < 0.01). Additionally, we find a shorter survival time for patients with LRP12 hypermethylation in a validation cohort (Fig. 2) as well as in the TCGA cohort for NSCLC (LUAD; Fig. 3) confirming the results.
Statistical modelling of whole-genome methylation enrichment experiments
Genome-wide enrichment of methylated DNA followed by sequencing (MeDIP-seq) offers a reasonable compromise between experimental costs and genomic coverage. However, the corresponding read-out of the experiments is qualitative only and quantification of the enrichment signals in terms of absolute levels of methylation requires specific transformation. We have developed a Bayesian statistical model that transforms the enrichment read counts to absolute levels of methylation and, thus, enhances interpretability and clinical application. We compared the method with competing models and show that the QSEA workflow outperforms other approaches and retrieves well-known lung tumor methylation markers that are causative for gene expression changes, demonstrating the applicability of QSEA for clinical studies. QSEA is implemented in R and available from the Bioconductor repository 3.4.
Dynamic adverse anti-cancer drug response at multi-omics levels
Anthracyclines are widely used as anti-cancer therapeutics despite of the fact that they lead to severe cardiotoxic effects in many patients. The mechanisms of anthracycline-induced cardiotoxicity are most likely multifactorial but remain largely unclear. In a European Consortium funded by Framework 7 we have analyzed human cardiac microtissues, mimicking essential physiological functions of the heart, and performed time-resolved measurements of most-widely used anthracyclines at the proteome, transcriptome and methylome levels. We identified essential molecular players of adverse drug response affecting sarcomere and mitochondrial functions. Furthermore, we agglomerated a large molecular interaction network from the ConsensusPathDB and used network propagation as concept for data integration and show that the overlay of multi-omics genome data onto molecular interaction networks increases the functional information compared to each single approach.
Genoytpe-phenotype correlations in type-2 diabetes mellitus
Previous genetic studies in patients have shown that multi-factorial disease phenotypes, such as diabetes mellitus (DM), cannot adequately be explained by a single or few variants but rather agglomerate contributions of multiple variants of genes organized in molecular networks. In this project we analyze patient variant profiles along with detailed phenotypic characterization of patients provided by the German Diabetes Cohort (GDC) Consortium at the level of biological networks in order to allow the differentiation of patients into subgroups according to secondary complications. These subgroups and comorbidities can be associated with characteristic genetic variant profiles and molecular interaction networks that determine the genotype-phenotype network of DM.
Machine learning methods for precision medicine
Machine learning approaches have emerged as the state-of-the-art methodology to infer predictions from large-scale data with numerous applications in science and economy. The group recently has successfully applied to the BMBF call on “Machine learning” with a research project that aims to develop novel machine learning methods and to apply these to the prediction of cancer therapy success in precision medicine. The goal of the project is to learn the sensitivity of the drug response of a biological system (cell line, patient tumors) from its molecular features and their relationships. Novel methods will develop constraints that allow better inclusion of background knowledge on biological pathways and gene-gene relationships.