gms | German Medical Science

Introduction: Many translational research projects in cancer research are utilizing genomic data directly from either patients or patient-derived models [1], [2], [3], [4]. Due to its sensitivity, the data needs to be stored in secured network segments, but must still be available for researchers and clinicians. Additionally, more and more genomic data is being used in patient care [1] e.g. for therapy boards. Therefore, it is a challenging but important task to enhance those projects by making this data available via ETL processes. In the Molecular Tumorboard (MTB)-Report project and the CRU5002 we aim to establish pipelines to integrate clinical and genomic data in data marts such as tranSMART [2] and cBioPortal [5], [6], Figure 1.

Methods: Across projects, we developed different ETL methods to extract and store genomic data. In the MTB-Report project, we import variant data directly from files on a molecular pathology server into the cancer documentation system Onkostar. The import is triggered manually in an Onkostar form, which sends the required patient information to a server in a secured network segment with reading access to the pathology server. It copies the corresponding files onto the server, stores the required extracted variant information in a CSV file, and imports it into Onkostar via a REST API. In the CRU5002 the FASTQ files are imported via a web interface into a CDSTAR [5] data lake instance and further created data are stored in a SEEK [7], [8] instance in a secured network segment to make the data Findable, Accessible, Interoperable and Reusable (FAIR [4], [9], [10]).

We frequently exported a subset of the Onkostar data to import it via a script into a tranSMART instance. During this process, we enriched the CRU5002 data with genomic data. Moreover, we are currently working on an import into our local cBioPortal instance.

Results: We imported clinical and genomic data from about 300 patients in MTB-Report as well as clinical data for 90 patients in the CRU5002. Since February 2022, we imported around 15 patients every two weeks into Onkostar. Before, this data had been inserted manually. Additionally, the data imported into Onkostar is further used in the MTB at the UMG. Moreover, it reduced the time for documenting the variants of a patient from an hour to minutes.

Discussion: Making the genomic and clinical data in the CRU5002 “FAIR” generates benefits for all researchers in the project.

Additionally, the automated import into Onkostar increased the data quality for the MTB, since the study nurse no longer needs to enter it manually. Alongside the actual data, we see the urgent need to add metadata about data sources, pipelines, and information models. One example was the export of analyzed panel data in the same data format whereas the change in the interpretation from using HG19 to HG38 led to huge differences [11].

Conclusion: Developing ETL processes enhances translational research projects by increasing the data quality and data provenance [4]. This improves results in the research projects, as well as in medical care for patients.

Funding: This work is partially funded by the VolkswagenStiftung within the MTB-Report project (ZN3424) and by the DFG within the CRU5002.

Acknowledgements: We would like to thank Kirsten Reuter-Jessen, Maximilian Papendick, Nicklas Lück, Elisabeth Heßmann and Marius Brunner for their support.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.