With the availability of more and more protein structures, structure-based design has become a key technology within the early phases of drug design. Protein structures are the only means by which a truly rational design approach gets in sight. While modeling techniques like docking and scoring dealing with small series of protein structures are well established, our methodologies to explore the wealth of information hidden in large collections of protein structures are still rather limited. Most search engines on protein structures are based on text rather than on structural elements, and the analysis of protein structure still requires labor-intense manual steps.

In this talk, new technologies will be presented addressing this opportunity to learn from large structure collections [see doi.org/10.1093/nar/gkx333 and doi.org/10.1016/j.jbiotec.2017.06.004]. On the one hand, the automation of structure preprocessing in the context of drug design plays a crucial role in exploiting large amounts of structural data. On the other hand, search methods allowing to perform geometric queries to structures enable knowledge-driven design decisions. Several examples ranging from interaction geometry analysis, molecular flexibility analysis to design by analogy will be presented.