Huber Group

Example image from a large-scale RNAi screen on populations of human cells stained for DNA (blue), tubulin (green) and actin (red). Images are automatically segmented, quantitative cell descriptors are computed and analysed for biological phenotypes by machine learning methods 

Dynamical modelling of protein life cycles from single cell fluorescence timer imaging

Independent filtering increases detection power for high-throughput experiments 

The Huber Group develops computational and statistical methods to design and analyse novel experimental approaches in genetics and cell biology. 

Previous and current research

Our aim is to understand biological systems through systems-wide maps and quantitative models. Our main tool is statistics - the science of computing with uncertainty, making rational inference based on incomplete and noisy data. Together with its sister discipline, machine learning, it helps humans to discover patterns in large datasets and to infer underlying mechanisms, predictive and causal relationships.

Our aim is biological discovery – to understand genetic and phenotypic variation between individuals on a genome-wide scale. We have projects in the areas of gene expression and regulation, in the genetics of complex phenotypes and genetic interactions, in cell division and cell migration and in cancer genomics.

The group brings together expertise from quantitative disciplines – mathematics, statistics, physics and computer science – and from different areas of biology to design and analyse genomic experiments and their biological interpretation. Computational and statistical methods are at the heart of systematic, large-scale experimental approaches. Our aim is to develop high-quality methods of general applicability that can be widely used in genomic research. We regard the publication of scientific software as an integral part of the publication of new methodical approaches and contribute to the Bioconductor Project.

Future projects and goals

An emphasis of the group’s work is on project-oriented collaborations with experimenters. We aim to develop the computational techniques needed to make new types of experiments feasible and to turn the data into biology. Among our current projects are:

• Large-scale systematic maps of gene-gene and gene-environment interactions by automated phenotyping, using image analysis, machine learning, sparse model building and causal inference.
• DNA-, RNA- and ChIP-Seq and their applications to gene expression regulation: statistical and computational foundations.
• Cancer genomics, genomes as biomarkers, cancer phylogeny.
• Image analysis for systems biology: measuring the dynamics of cell cycle and of cell migration of individual cells under normal conditions and many different perturbations (RNAi, drugs).
• Systematic mapping of molecular interactions and life cycles within single cells.
• Open source software for genomics, high-throughput phenotyping and statistical bioinformatics, to support reproducible research and wide dissemination of state-of-the-art methods.