Design, Selection, and Evaluation of a General Kinase‐Focused Library

Abstract

The human kinome [1] is a highly conserved target family composed of more than 500 different proteins. Kinases play fundamental roles in many intracellular pathways such as cytokinesis, cell proliferation, differentiation, and apoptosis, and are therefore implicated in various diseases. Currently there are significant efforts in therapeutic areas such as cancer and inflammation to identify novel kinase inhibitors. Random screening of a discovery compound library often results in a hit rate of 0.1 %, [2] whereas focused library screening could improve this rate to ! 1 %. Consequently, libraries focused toward kinases have become starting points in screening campaigns [3] and are complementary to conventional high-throughput screening (HTS) [4] of discovery libraries. Compound collections catered to target families [5] such as kinases present a unique opportunity to explore discrete chemical, biological, and property spaces. In contrast, HTS libraries are built to represent maximum diversity in the chemical and biological properties of compounds. Focused libraries are also appealing due to decreased synthesis, repository management, and screening costs. The present work describes a rapid computational process to select a collection of compounds targeted as kinase inhibitors, combining the advantages of 2D and 3D virtual screening methods for use in kinase-focused screening campaigns.In designing a general kinase screening library, the challenge lies in defining the chemical and biological space that identifies compounds with utility against any of the many possible kinase targets. Whereas some inhibitors such as staurosporine are known to be active against many kinases, others show a specific inhibitory profile. [6,7] The limited selectivity of many inhibitors is due to the fact that the catalytic ATP binding domain targeted is highly conserved. In recognition of the difficulty of designing selective kinase inhibitors, dual-and multitarget kinase inhibitors were developed.[8] Such "dirty drugs" (i.e., sorefenib, sunitinib) have gained much attention in recent years and show potential to be advantageous in cancer therapy. These clinical developments have also contributed to the increased interest in general kinase inhibitor libraries. [9] In this work, our goal was to develop a focused library selection procedure based on a 2D similarity search combined with 3D target-based filtering. A recent review argues that "2D fingerprints are surprisingly effective in many search situations in comparison with more complex 3D designs".[10] Indeed, 2D approaches allow a rapid analogue search from various databases.[11] We find the combination of 2D and 3D methods has distinct advantages; it can decrease the number of false negatives, and 2D methods can represent a pre-filtering tool that enables real-time 3D virtual screening using traditional docking algorithms tailored to the evaluation of large numbers of molecules. Herein we describe our methods and characterize the general kinase-focused library.

Rapid Identification of Potential Drug Candidates from Multi-Million Compounds’ Repositories. Combination of 2D Similarity Search with 3D Ligand/Structure Based Methods and In Vitro Screening

“…It is widely discussed that new chemotype discovery or scaffold hopping is possible if we go below T2D = 0.65 similarity value [ 56 , 57 ]. Lower similarity thresholds also afford higher structural diversity, which might lead to novel chemotypes with or without scaffold hopping even though the hit rate is somewhat reduced [ 58 ]. It is important to note that while the Tanimoto coefficients of using chemical hashed and related connectivity-based fingerprints are comparable (as used in all cases of the present account), substructure or pharmacophore-based fingerprints require different thresholds for similarity selection.…”

Combination of Pharmacophore Matching, 2D Similarity Search, and In Vitro Biological Assays in the Selection of Potential 5‐HT6 Antagonists from Large Commercial Repositories

“…Cluster 1 contains 12 seed molecules (seed numbers : 2, 18, 19, 20, 24, 25, 28, 29, 30, 40, 43, 44) and 5 hits (hit numbers: 1, 2, 4, 9, 6), Cluster 2 (seed numbers: 1,5,11,12,14,22,31,33,34,35,48) had 11 seeds and 6 hits (hit numbers 7, 8, 5, 10, 12, 13), while cluster 3 (sulfon-bridge) contained only 12 seed molecules (seed numbers: 3, 4, 6, 23, 27, 36, 37, 41, 42, 46, 47 and 49) and no molecules from the first round hits. (The structure of the seeds is displayed in the Supplementary Materials).…”

“…In the meantime, various virtual screening methods have become available for the selection of focused libraries from such multimillion compounds repositories . Focused library screening could improve the hit rate to ≥1%, while random screening of a discovery library often results in a hit rate ≤ 0.1% . Furthermore, in many cases, virtual screening methods and in vitro HTS were combined and found complementary to each other .…”

Property-based characterization of kinase-like ligand space for library design and virtual screening

“…Various approaches have been applied previously to assemble kinase-focused compound libraries (virtual and physical as well), including substructure-based methods [8][9][10][11][12][13] and similarity-based methods. [14][15][16] Most recently, Singh and coworkers explored the possibility of characterizing kinaselike ligands based on physicochemical descriptors. 17 With the increasing amount of publicly available inhibitor activity data, 18 this approach becomes an attractive opportunity, since substructure-and similarity-based methods inherently retrieve molecules that are structurally similar to the reference compoundIJs), limiting the ability to identify inhibitors with novel scaffolds.…”