Kinases are an important class of drug targets since their dysregulation can cause severe diseases such as cancer, inflammation, and neurodegeneration. Finding selective drugs, however, is challenging due to the highly conserved binding sites across the roughly 500 human kinases, which can lead to unwanted side-effects. The underlying off-targets are often not trivial to predict or to explain from a sequence-based perspective.
In our KiSSim project, we explored kinase similarity from a physicochemical and structural point of view. We thereby introduced a kinase-focused and subpocket-enhanced fingerprint to compare kinase pockets across the structurally covered kinome.
Take a look at our ChemRxiv preprint “KiSSim: Predicting off-targets from structural similarities in the kinome” to find out more.
We publish our code as
kissim Python package available at GitHub and as
conda package (check out the
kissim documentation for more details). All data and analyses are integrated in the
kissim_app GitHub repository.
Thanks to all co-authors for working together on realizing this project: Dominique Sydow, Eva Aßmann, Albert Kooistra (University of Copenhagen), Friedrich Rippmann (Merck Healthcare KGaA), and Andrea Volkamer.