Biostatistics/Bioinformatics, Pharmacometrics and Artificial Intelligence

The Biostatistics and Bioinformatics Core Facility provides expert guidance and support for processing, preparing, and analyzing a wide range of biological, clinical, and pathological data. The facility utilizes established, standardized workflows for processing next-generation sequencing data, including pipeline solutions from the nf-core community project, ensuring efficient and reproducible handling of biomaterial data. Services include experiment planning (e.g., sample size estimation), selection of appropriate technical methodologies, and advanced biostatistical and bioinformatics analyses. The team is skilled in managing high-dimensional and heterogeneous datasets using scripting languages (e.g., R, Python) and scalable computing resources. Analytical approaches encompass supervised methods (e.g., regression, classification, survival analysis) and unsupervised techniques (e.g., clustering, dimensionality reduction), along with deep learning methods (e.g., convolutional neural networks) for image analysis, including tasks such as feature extraction, segmentation, and classification in medical and biological imaging.

Deep-learning techniques are used for different tasks in computational pathology and in the analysis of sequencing and multiplex immunofluorescence data. For the latter, the core facility established robust workflows for quality control, processing and analysis, including deep-learning-based approaches for cell segmentation and unsupervised and supervised cell phenotyping. Interpretable convolutional motif kernel networks, developed in a collaborative project, are applied for robust prediction of drug response phenotypes using sequencing data. In computational pathology, established frameworks for the analysis of hematoxylin and eosin (H&E) stained whole-slide images were customized to our specific datasets and research questions, encompassing cell core segmentation, tissue classification, and the extraction of (computational) histopathological features for downstream analysis.