The combination of "liquid biopsies", machine learning and data visualization will allow earlier and more accurate prediction of cancer relapse.
In modern tumor diagnostics, quantitative markers from computer-aided image analysis, as well as molecular multi-scale analyzes ("OMICS") have already demonstrated high diagnostic value, e.g. for determining recurrence probabilities. The trend is currently toward low-invasive procedures: liquid biopsies that use blood or bone marrow to extract tumor cells or cell-free DNA. A combination of genome-wide DNA analysis with mRNA expression data and robust image features could significantly improve the accuracy of recurrence predictions, as compared to unimodal analyzes to find previously unknown correlations between data sets / patients. Despite intensive research in this field, currently there are no appropriate methods and platforms for merging image and OMICS data as a prerequisite for the development of certified, clinical analysis workflows. Especially in rare diseases, there is also the problem of incomplete data sets and a small number of samples.
The innovation of VISIOMICS lies
- in combining and developing methods that enable the preparation, integrative analysis and visualization of multimodal data sources in rare cancers using metastatic neuroblastoma as an example
- in liquid biopsies, which in unimodal analysis already allow a more precise prediction of the probability of recurrence than the primary tumor and
- in the development of modules that can be promptly certified as clinical workflows.
To achieve this goal, data sets from clinical, image (multi epitopeligand cartography as well as iFISH) and OMICS data (HD-SNP array, RNA sequencing) from liquid biopsies (disseminated tumor cells) are progressively incorporated, adding expert knowledge and use of bioinformatic and machine learning and visual analytics methods. The interface between the lab, bioinformatics method development and tools for data integration, visualization and user interaction is an established platform that provides a central database and interfaces for certifiable, clinical analysis workflows.
Project role VRVis
The data that form the basis of the research within the framework of the VISIOMICS project are extremely heterogeneous and complex. Recognizing relationships between these data is the first step in building models that allow tumor diagnosis to be refined. In the VISIOMICS project, VRVis is a partner for the development of web-based user interfaces and interactive tools for data visualization and exploration in order to accelerate hypothesis formation and knowledge gain based on existing patient studies. The quantitative visual analytics solutions are developed in dialogue with the VISIOMICS partners and closely linked to the planned automated analysis tools to best support the complex analytical workflows.