gms | German Medical Science

Introduction: Microbes have an influence on the outcome of cancer therapy. Despite this evidence, the number of microbes known to directly cause the development of cancer remains small, making further research necessary [1]. Within the context of the KFO5002, we loaded buccal and rectal samples from pancreatic cancer (PDAC) patients into our clinical data warehouse, cBioPortal, which enables holistic analyses on the cancer patient [2], [3]. The data derived from the gastroenterology department of the University Medical Center Göttingen and were stored as taxa counts. While the buccal samples were extracted by marker gene analysis, the rectal samples were extracted by whole metagenome analysis. Comparable research in this area is Poore et. al. who performed metatranscriptomic analysis to determine the microbial composition and loaded the taxa counts on a log2-scale in cBioPortal [4].

Methods: We developed ETL methods to extract and store microbiome and clinical data. Depending on the type of analysis, the data was available in a Phyloseq object in R, requiring the use of an R wrapper in our Python-ETL-script. The clinical data included in the Phyloseq object were comprehensive and provided information on, e.g., the patient's therapy, demographic details, operations or previous diseases of the patient. To annotate the clinical data, we used the suggested column names and descriptions from the clinical data dictionary of the OncoTree api recommended by cBioPortal. Since cBioPortal demands microbiome data on a logarithmic scale as described by Poore et. al. we had to transform the taxa counts. We used a transformation from the R package metagenomeSeq that performed log2(x+1) with taxa counts. We then loaded the microbiome data into cBioPortal and analyse it.

Results: We were able to load microbiome data and clinical data from 59 PDAC and 49 chronic pancreatitis (CP) patients into cBioPortal. The cBioPortal incorporated volcano plot does not reveal any differential abundant taxa between both groups regarding the buccal and rectal microbiome. That is contrary to the previous analysis in which differential abundances were obtained by LefSe [5]. Hereby, several taxa were differential abundant in rectal samples between CP and PDAC groups.

Discussion: The exploratory analysis techniques of cBioPortal allowed researchers to see their microbiome data from a different perspective. The group-comparison function and filtering of patients in cBioPortal were found to be useful. While the automatic generation of statistical plots such as p-value, Q-value or Volcano Plots saves time and benefits analysis workflows, it is nevertheless necessary for further microbiome analyses in cBioPortal to be able to use further customised statistical approaches (e.g. Mann-Whitney-U-Test, Alpha-Diversity, Beta-Diversity).

Conclusion: Integrating microbiome data and clinical data in cBioPortal makes data more accessible and discoverable for the clinicians, but lacks statistical tools for microbiome analysis in cBioPortal. Therefore, future research can work on implementing statistical tools for microbiome analyses in cBioPortal.

Funding: This work is funded by the DFG within CRU5002 (Project 426671079).

Ethics: The authors declare that a positive ethics committee vote has been obtained. All work involving patient data is carried out under ethics vote (11/5/17).

Acknowledgements: We would like to thank Prof. Dr. rer. nat. Andre Franke for his support.

The authors declare that they have no competing interests.