Meeting information

Computer Aided Drug Design - Gordon Research Conference

Integrating Big Data and Macromolecular Protein Structures into Small Molecule Design

Date: 14 - 19 July 2019

Location: West Dover, VT, US

Agenda: View the scientific programme here.

Peter Schmitdke and Daniel Alvarez-Garcia will both present a poster at the event. Check out the abstracts here:

Redefining the Kinome Tree – Local Sequence Based Selectivity Analysis of the Human Kinome 

Authors: Peter Schmidtke 

 The human kinome is a large class of structurally similar proteins with very interesting therapeutical implications. Drug discovery on human kinase targets is always accompanied by selectivity issues which usually require several compound optimization rounds to improve the selectivity towards your target of interest. In the past, several different binding sites have been used to alter kinase activity. The ATP binding site and the prolonged type 2 and type 3 subpockets are still the most often targeted sites today.  

The Manning et al. Kinome tree [1] was published in 2002 to shed some light into the kinase protein family and study their phosphorylation properties. It became during the years an important reference in drug design to map known kinase inhibitors and evaluate their selectivity. However, as pointed out by several recent contributions [2-4] the initial tree was built using a global sequence alignment and comparison amongst all human kinases.  

Here we present our work on deriving multiple binding site specific kinome trees using locally refined sequence alignments. These focused trees are better suited for drug design projects to evaluate selectivity challenges one might face during inhibitor development against a specific pocket of a particular kinase. 

From a kinase domain specific alignment, we use reference structures of known kinases, known and putative binding sites identified by fpocket in 3decision® and a modified substitution matrix to derive these binding site specific kinome trees.  

The potential of our binding site centered kinome trees is demonstrated by comparison to known inhibitory activity data and to the initial human kinome tree proposed by Manning et al. We believe that these new kinome trees can be used to assess early-on which kinases to closely monitor for selectivity issues and help the scientist choose the most strategic binding site to target.   

[1] Manning et al. Science. 2002 Dec 6;298(5600):1912-34. 
[2] Hanson et al. Cell Chem Biol. 2018 Dec 11. pii: S2451-9456(18)30412-4 
[3] Kooistra & Volkamer, Ann Rep Med Chem, 2017, 263-299 
[4] Christmann-Franck et al. JCIM, 2016, 56(9) 

Drug design in a local environment: Sub-pocket similarity search 

Authors: Daniel Alvarez-Garcia, Peter Schmidtke 

The identification and optimization of molecules that bind specifically to a binding site is the central work in any structure-based drug discovery project. In this challenging task we usually want to extend the ligand into an unoccupied space within the pocket or to replace a fragment to optimize the binding affinity. For this, the drug designers typically rely on their knowledge in medicinal chemistry and on classical modeling techniques. To speed up the ideation process, the information found in experimentally resolved protein structures could be harvested and used. However, a systemic and large-scale structural analysis of this kind can be difficult and time-consuming to set up. 

Here we present a new large-scale sub-pocket search methodology that will allow us to find sub-pockets with similar local environments within all 3D structures available in the public and private domain (RCSB PDB and internal structures). The method is simple to use and can be applied to gather potential binders and scaffolds for binding sites of interest using all PDB structures available today.