Novel Kinase Inhibitors by Reshuffling Ligand Functionalities Across the Human Kinome

Protein kinases remain among the
most versatile and prospective
therapeutic drug targets with currently 15 distinct compounds approved
for use in humans and numerous clinical development programs. The
vast majority of kinase inhibitors bind at the ATP site. Here we present
an integrated workflow to amplify the rapidly increasing space of
structurally resolved small molecule kinase ligands to generate novel
inhibitors. Our approach considers both receptor-based similarity
constraints in cocomplexes and ligand-based filtering/refinement methods
to generate novel, drug-like matter. After building a comprehensive
database of the structural kinome and identifying ATP-competitive
ligands, we leverage local site similarities and site alignments to
shuffle ligand fragments across the kinome. After extensive curation
and standardization, our automated protocol starting from 936 cocrystal
ATP-competitive binding sites generated about 150 000 new ligand
structures among them over 26 000 lead-/drug-like compounds;
the majority of those are novel based on structural similarity and
scaffolds. In a retrospective analysis we demonstrate that our protocol
produced known potent kinase inhibitors and we show how docking can
be applied to prioritize the most likely efficacious compounds. Our
workflow emulates a common strategy in medicinal chemistry to identify
and swap corresponding moieties from known inhibitors to generate
novel and potent leads. Here, we systematize and automate this approach
leveraging available knowledge covering the entire human Kinome.

蛋白激酶（Protein kinases）仍是目前应用最广泛且最具前景的治疗性药物靶点，目前已有15种不同的相关化合物获批用于人类治疗，另有大量项目处于临床开发阶段。绝大多数激酶抑制剂均结合于ATP结合位点（ATP site）。本文提出一套整合性工作流程，可扩增快速增长的已解析结构小分子激酶配体库，以生成新型激酶抑制剂。本方法同时兼顾共结晶复合物（cocomplexes）中基于受体的相似性约束，以及基于配体的筛选/优化手段，以生成具备类药性质的新型化合物。在构建完整的结构激酶组（kinome）数据库并识别ATP竞争性配体后，我们利用局部位点相似性与位点比对方法，在整个激酶组范围内对配体片段进行重排。经过全面的整理与标准化流程后，我们的自动化方案从936个共结晶ATP竞争性结合位点出发，生成了约150000个全新配体结构，其中包含超过26000个先导化合物/类药化合物；其中绝大多数基于结构相似性与分子骨架来看均为新型分子。在回顾性分析中，我们证实该方案可生成已知的强效激酶抑制剂，并展示了如何通过分子对接（docking）对最具潜在疗效的化合物进行优先级排序。本工作流程模拟了药物化学（medicinal chemistry）领域的常规策略：从已知抑制剂中提取对应官能团片段并进行交换，以生成新型强效先导化合物。本文将该方法进行系统化与自动化处理，依托覆盖整个人类激酶组的现有知识完成开发。