Personalized Medicine Research

Precision oncology is a treatment strategy in cancer medicine that establishes patient specific targeted therapies based on genomic and epigenetic profiles. It is focused on revealing molecular mechanisms that contribute to initiation and progression of the cancer, and identification of the clinical implementations based on actionable variants. Within this framework, we aim to develop a clinical decision support system which makes all data relevant to therapeutic decisions available to Molecular Tumor Boards (MTBs). We are focused on development and integration of analysis pipelines for molecular tumor table backends mainly centered around high-throughput data extraction and integration, automated processing of incoming data, annotation of therapeutic options and visualisation of network-derived contexts and analysis results.

Clinical Variant Annotation Pipeline (ClinVAP)

Clinical application of precision oncology necessitates the employment of fast, robust and automated pipelines to process genomics data and retrieve the mutational profile of the patients. To meet this need, we developed ClinVAP which translates the long list of variants to a clinically relevant and structured information. Based on somatic single nucleotide variants, ClinVAP reveals the molecular mechanisms initiating carcinogenesis and actionable variants. Annotated, filtered and prioritized variants along with the therapeutic suggestions are presented to MTBs as a clinical decision support mechanism.

The cellular machinery is comprised of many individual metabolic, regulatory and signalling events. Put together they build vast networks of interconnected reactions, interactions and relations. Our group researches various methods to leverage the information in these networks as well as to analyze clinical data in that context for additional insight.

DeRegNet

We develop DeRegNet, a software to find deregulated subnetworks in large directed biological networks based on scores that were derived from various omics data. The underlying method relies on fractional integer programming to arrive at optimal solutions. This open-source software allows many flexible use cases for functional annotation and hypothesis-generation for omics based research.

Biological Networks in Graph Databases

To utilize the analytical power of graph databases we are developing an application to persist biological models available in the SBML format (http://sbml.org/Main_Page) in graph databases. The resulting graph representations allow the creation of network mappings (e.g. regulatory networks, metabolic networks and others) which can be exported for use in other applications but also be traversed with graph algorithms on the database level. This grants performant access to network contexts around genes and their products, pathfinding for drug target searches up- or downstream of genes of interest and many more applications.

Subnetwork-Guided Analyses

By making use of the integrated systems biology knowledge, the deregulated subnetworks generated by DeRegNet allow us to explore putative disease mechanisms and relations that go beyond already disseminated pathway knowledge. Thus, one of our current focal points is to assess how to best utilize these subnetworks for Machine Learning applications in personalized medicine, such as disease prediction or patient stratification.

Visual Analytics

To prevent, diagnose and treat disease, personalized medicine uses information about a person’s genes, proteins, and environment. Instead of poring over spreadsheets or reports, it is easier to use charts and graphs to visualize these large amounts of data as the human brain is trained to see and understand trends, outliers, and patterns in data visually. Therefore, we are working on new ways to visually include patient data to the MTBs. Our first step is to integrate an interactive visualization of the DeRegNets into the MTB to show genes that are possibly related to a specific cancer or may be an unknown target for drug treatment.