inSili.com LLC, Segantinisteig 3, 8049, Zurich, Switzerland.
Department of Chemistry and Applied Biosciences, Swiss Federal Institute of Technology (ETH), Vladimir-Prelog-Weg 4, 8093, Zurich, Switzerland.
Angew Chem Int Ed Engl. 2017 Sep 11;56(38):11520-11524. doi: 10.1002/anie.201706376. Epub 2017 Aug 7.
Drug discovery is governed by the desire to find ligands with defined modes of action. It has been realized that even designated selective drugs may have more macromolecular targets than is commonly thought. Consequently, it will be mandatory to consider multitarget activity for the design of future medicines. Computational models assist medicinal chemists in this effort by helping to eliminate unsuitable lead structures and spot undesired drug effects early in the discovery process. Here, we present a straightforward computational method to find previously unknown targets of pharmacologically active compounds. Validation experiments revealed hitherto unknown targets of the natural product resveratrol and the nonsteroidal anti-inflammatory drug celecoxib. The obtained results advocate machine learning for polypharmacology-based molecular design, drug re-purposing, and the "de-orphaning" of phenotypic drug effects.
药物发现受寻找具有特定作用模式的配体的愿望所驱动。人们已经意识到，即使是指定的选择性药物，其大分子靶点也可能比人们通常认为的要多。因此，在设计未来药物时，必须考虑多靶点活性。计算模型通过帮助在发现过程的早期消除不合适的先导结构和发现不期望的药物作用，协助药物化学家进行这项工作。在这里，我们提出了一种简单的计算方法来寻找药理活性化合物的以前未知的靶标。验证实验揭示了天然产物白藜芦醇和非甾体抗炎药塞来昔布的以前未知的靶标。所获得的结果支持基于多药理学的分子设计、药物再利用以及表型药物作用的“去神秘化”的机器学习。
