Read-Across the Targetome - An integrated structure- and ligand-based workbench for computational design of novel tool compounds

如何探测和验证潜在的途径或靶点仍然是生命科学基础研究的关键问题之一。这些研究通常缺乏合适的化学工具化合物来阐明特定蛋白质的功能。因此，像结构基因组学联盟这样的大型联盟已经形成，通过经典的化学合成和广泛的蛋白质结构测定工作来生成用于验证生物靶点的工具化合物。虽然这些联盟将继续专注于他们的实验方法，但跨目标组读取项目将为生成一套针对新目标的综合工具化合物提供计算解决方案。该项目假设基于相似性原理（相似的口袋结合相似的化合物），最终目标是利用蛋白质口袋相似性从一个目标推断化合物信息到另一个目标。有了这个概念，研究者试图回答核心问题：结合位点相似性可以用来提出新的目标工具化合物吗？基于这种新的刀具复合识别范式，将开发一个整体工作台。目前在蛋白质数据库（PDB）中有超过127,000个大分子结构，ChEMBL中有超过168万个已测试的化合物，丰富的结构和结合数据是免费提供的，这里将系统地探索这些数据。首先，从PDB中收集蛋白质结构。对于每个蛋白质的所有结构，潜在的口袋将被识别并聚集到集合口袋中。接下来，将开发一种新的高效的基于结构的结合位点比较算法，以寻找最相似的口袋，考虑到蛋白质在集成方面的灵活性。配体和结合数据将从ChEMBL中提取，过滤并分配到各自的口袋中。利用新算法，已知的与检测到的邻近口袋结合的配体可以被阐明。这些化合物可以作为功能注释的化学探针或作为虚拟筛选的新型聚焦化合物库。整个过程将以网络服务的形式提供，将数据和方法整合到一个地方，并允许来自不同生命科学学科的研究人员简化访问方法。结合利用现有的结构和化学信息，以及从已知化合物到新的类似结合口袋的知识转移，将有助于加快途径和靶点验证领域的生物和制药研究。

How to probe and validate a potential pathway or target remains one of the key questions in basic research in life sciences. Often these investigations lack suitable chemical tool compounds for the elucidation of the function of a specific protein. Therefore, large consortia, like the Structural Genomics Consortium, have formed to generate tool compounds for the validation of biological targets via classical chemical synthesis and extensive protein structure determination efforts. While these consortia will continue with their focused experimental approaches, the read-across the targetome project will offer a computational solution for the generation of a comprehensive set of tool compounds for novel targets. The project hypothesis is based on the similarity principle (similar pockets bind similar compounds) with the ultimate goal of using protein pocket similarity to extrapolate compound information from one target to another. With this concept, the investigator seeks to answer the central question: Can binding site similarity be used to propose tool compounds for novel targets? Based on this new paradigm for tool compound identification, a holistic workbench will be developed. With currently over 127,000 macromolecular structures in the protein data bank (PDB) and over 1.68 million tested compounds in ChEMBL, a wealth of structural and binding data is freely available, which will be systematically explored here. First, protein structures will be collected from the PDB. For all structures per protein, potential pockets will be identified and clustered to ensemble pockets. Next, a novel efficient structure-based binding site comparison algorithm will be developed to find the most similar pockets considering protein flexibility in terms of ensembles. Ligand and binding data will be extracted from ChEMBL, filtered and assigned to the respective pockets. Using the novel algorithm, ligands known to bind to the detected neighboring pockets can be elucidated. These compounds can be selected as chemical probes for functional annotations or as novel focused compound libraries for virtual screening. The whole procedure will be made available as a web-service to combine the data and methods in one place, as well as to allow simplified access to the methodology for researchers from different life science disciplines.The combined usage of available structural and chemical information, as well as knowledge transfer from known compounds to novel similar binding pockets, will help to speed-up biological and pharmaceutical research in the area of pathway and target validation.