Computational Drug Discovery

Research in this group, headed by Gerard van Westen, focusses on computational methods integrated in different parts of the drug discovery process. More specifically, topics include innovative treatments for cancer, selectivity modeling, translational research, allosteric modulation, drug resistance modeling, and novel methods in the domain of machine learning. Computational approaches have integrated into science over the last decades as they have in our daily lives. Computers are invaluable in the complex process that drug discovery represents, ranging from playing a supportive role to research projects being run fully in silico. In fact, the term in silico has been added to the more traditional terms in vitro and in vivo. Gerard van Westen has over a decade of experience in computational drug discovery both in a commercial and academic context.

For our research we explore diverse data sources and diverse methods from the domains of (computational) chemistry, bioinformatics, and computer science. All research in the group can be divided into two key research areas: data mining (statistical modeling) and structure-based drug discovery methods (SBDD or molecular modeling).  With respect to statistical modeling, such methods rely on the availability of (large amounts of) prior data. Algorithms are used to leverage this data in the context of scientific research. In practice this typically means making predictions about the biological potential of novel candidate drugs (i.e., will it bind to a required target, will it avoid binding to an off-target), or trying to understand why a given candidate drug exerts an observed phenotype (e.g., mode-of-action elucidation) Public resources which are heavily used include the ChEMBL database, Uniprot, ZINC, and many others.

With respect to molecular modeling, methods typically require the availability of a crystal structure or high-quality homology model. When this type of atomic resolution data is available for a given project, we apply methods such as docking (fitting a small molecule in a binding site), molecular dynamics (studying how atoms interact at a nanosecond timescale), and free energy perturbation methods (quantifying small molecule affinity for a target using molecular dynamics). Public resources used include the protein data bank (PDB), directory of useful decoys enhanced (DUDe), ZINC, and several others.

Both disciplines are investigated in the context of fundamental research, development of new algorithms and understanding algorithm function, and as applied techniques, using existing methods to answer scientific questions in the context of another research project. The applied techniques are usually done in collaborative projects such as the Leiden based Data-Driven Drug Discovery Network (D4N) or in projects that are executed as part of a collaboration with a commercial partner.

In principle our methods can be applied to any family of proteins or any disease (target and disease agnostic). However, over the years a key expertise in certain niches was developed. For protein families this expertise includes G Protein-Coupled Receptors (GPCRs) and kinases, and for disease areas this is mainly in the context of cancer research. For more information and a better understanding, why don’t you explore one of the more detailed pages listed under the research projects. Some typical techniques that are under extensive investigation are proteochemometrics in the case of statistical modeling and free energy perturbation in the case of molecular modeling.