In the bioinformatics group at the Institute of Environmental Medicine and Integrative Health, we are using statistic and computational approaches to understand biological data in the broader context of environmental diseases. We are analysing and combining multidimensional data from immunology, microbiology and clinical studies to identify biomarkers for diagnosis and personalized therapies. Our work mainly focuses on the course of viral infections (Covid-19) and human-bacteria interactions in the microbiome of allergic diseases (Atopic Dermatitis). Apart from clinical questions, we also review software tools and develop new methods to improve data processing and analysis.

Microbiome in clinical studies

It is long known that the human microbiome – i.e. the microbes living in and on our body surfaces – is associated with human health. With the development of DNA sequencing technologies, microbiome research has been gaining momentum, as it identifies millions of microbes from hundreds of samples at the same time. After thoroughly pre-processing these gigabytes of high-dimensional data, we can associate the microbiome with patient characteristics such as age, or find different proportions of harmful and beneficial bacteria in response to the dosage of a new drug (Figure 1).

Improving microbiome data quality

While it is established that the human microbiome is altered in diseases like Atopic Dermatitis or Inflammatory Bowel Disease, the microbiome is far from being used as a precise clinical biomarker. In order to find robust clinical associations, pre-processing and quality control of microbiome data is essential. The complex process from sample handling to analysis leads to specific errors and biases associated with microbiome data (e.g. DNA sequencing errors and differences in cell lysis by bacterial cell wall structure). These errors and biases make it challenging to distinguish real biological findings from experimental noise. We tackle these key challenges of microbiome research with benchmarking laboratory and computational methods, and by developing new algorithms for microbiome data processing. Our novel tools for getting strain-level taxonomic information, and for removing contaminants from demanding low-biomass samples like skin, are central steps towards high-quality microbiome data. Our final goal is to generate robust and applicable microbiome data, which can be used to improve patient diagnosis and care.