Computational oncology & molecular modeling (SIB)

Our main research focus is personalized oncology. Our aim is to make use of all available –omics data to guide treatment decisions for all oncology patients. We are developing baseline and on treatment predictive biomarkers in order to determine the optimal treatment sequences.

The following projects are conducted in close collaboration with Zoete Lab.

Personalized oncology

Selecting the best treatment option for patients based on large –omics data sets require cutting edge machine learning algorithms. We are actively developing methodologies to provide interpretable machine learning results that pinpoint to the essential aspects leading to the proposed classification by the algorithm. In addition, we are also developing techniques to make use of the existing body of knowledge (Pubmed, Onco-KB, TCGA, …) to prioritize some branches of the decision trees. Some of these developments are conducted in collaboration with the Swiss Data Science Center (SDSC).

Structure-based drug design

As part of a long-lasting interest of our group in structure based drug design, we are pursuing several projects to design high affinity small molecule inhibitors for important targets in oncology and, most importantly, immuno-oncology. Current projects include IDO for which several nano-molar compounds have been generated as well as STING. Our program starts with computer aided drug design, organic synthesis, soluble and cellular tests, all the way to proof of concept studies in mouse models.

Our algorithms are now being used for personalized oncology, where the impact of somatic mutations on specific targeted therapies is evaluated. Recently, we have launched the Swiss Personalized Oncology project that aims at providing a seamless web interface to access these tools for the clinicians making decisions within molecular tumor-boards.

Structure-based protein design

Structure-based protein design has also been a strong interest of our group. We use molecular dynamics and free energy simulations to study key proteins like the TCR-Peptide-MHC complex and to improve relevant biological properties such as the on-rate, off-rate or the affinity. Since our approach is based on the laws of physics, its universal nature allows direct application to other proteins or complexes of interest like, for example, signaling domains.

We are now using these methods for personalized oncology by studying the structural conformation of neo-antigens presented within the MHC context or the structural TCR repertoire selected against such neo-antigens.