Our research focus is on gene regulation. Transcription factors are the molecules which decode the regulatory instructions encoded in the DNA by binding to specific locations in the genome. Through the analysis of public data, we are investigating the molecular processes that recruit transcription factors to their target DNA sites in both a tissue and development stage-specific manner. Of particular interest to us are those processes which cause disease if disrupted.

Novel algorithms for analysing ChIP-seq and other chromatin profiling data.

We have introduced and are now working on further extensions of “probabilistic partitioning”, a flexible method for discovery, characterization and alignment of chromatin patterns. In collaboration with Bernard Moret at EPFL; we have modified tree-inference algorithms from phylogeny in such a way that they can be applied to cell differentiation trees in conjunction with chromatin profiling data.

Inference of transcription factor binding specificity from high-throughput data.

Some time ago, we introduced the first high-throughput SELEX protocol with a yield of more than 10,000 binding sequences per transcription factor (collaboration with Nicolas Mermod from UNIL). Ever since then, we have worked on the computationally challenging problem of inferring accurate binding site models from high-throughput data obtained with a variety of technologies, all of which are prone to artefacts and biases. In 2010, we participated in the TF-DNA motif recognition challenge organized by DREAM, finishing in an honourable second place. Currently we apply our experience and know-how in this area to data on heteromeric transcription factors generated with novel technology in Bart Deplancke’s lab at EPFL.

Characterization of ultraconserved non-coding elements (UCNEs) and genomic regulatory blocks (GRBs).

UCNEs are the most conserved DNA sequences in vertebrates. Yet their molecular and cellular function remains completely enigmatic. UNCEs occur as large clusters in the genome, so-called GRBs, which typically surround key developmental genes. By analysing the fate of UNCEs after whole genome duplication in the fish lineage, we were able to show that UCNEs of the same GRB act in a highly cooperative manner. Our current work focuses on the evolutionary mechanisms involved in the creation of UCNEs and the expansion of GRBs.

Using molecular profiling data for molecular diagnostics.

We are developing methods for automatic classification and diagnosis of molecularly profiled clinical samples. Specific problems we addressed in the past include sub-classification of lung cancers based on gene expression profiles and diagnosis of acute myeloid leukaemia (ALM) with the aid of flow cytometry data. To compare our methods with those of the world leading teams, we have participated in several open challenges organized by the DREAM and sbv Improver consortia. Currently, we are evaluating the potential value of ChIP-seq data obtained from surgically removed tumours or biopsies for prognosis and treatment response prediction.

More information about our research