Torres-Urtizberea Endika, Borrell José I, Puig de la Bellacasa Raimon, Estrada-Tejedor Roger
Grup de Química Farmacèutica, IQS School of Engineering, Universitat Ramon Llull, Via Augusta 390, E-08017 Barcelona, Spain.
Int J Mol Sci. 2025 May 7;26(9):4457. doi: 10.3390/ijms26094457.
In the classic medicinal chemistry hit discovery procedure, large virtual libraries undergo different filtering and prediction steps until a small group of molecules is selected for their subsequent synthesis and biological testing. The starting molecular libraries can easily be composed of millions of molecules, hindering the selection of the most representative and promising compounds. Moreover, the resulting molecular systems tend to be overcomplex structures, hardly attainable, and often involve extrapolations of the prediction models used. We present a rational-based method to reduce the structural complexity of molecular candidates without compromising their biological activity, improving the attainability and efficiency of hit discovery. This approach has been successfully applied to identify potential tyrosine kinase dual inhibitors against Fibroblast Growth Factor Receptor 2 (FGFR2) and Insulin-Like Growth Factor 1 Receptor (IGF1R), a set of overexpressed proteins in different cancers, such as pancreatic ductal adenocarcinoma (PDAC).
在经典的药物化学先导化合物发现过程中，大型虚拟库要经过不同的筛选和预测步骤，直到选择一小部分分子进行后续合成和生物学测试。起始分子库很容易由数百万个分子组成，这阻碍了最具代表性和前景的化合物的选择。此外，所得分子系统往往是过于复杂的结构，难以实现，并且常常涉及所用预测模型的外推。我们提出一种基于理性的方法，在不影响分子候选物生物活性的情况下降低其结构复杂性，提高先导化合物发现的可实现性和效率。该方法已成功应用于鉴定针对成纤维细胞生长因子受体2（FGFR2）和胰岛素样生长因子1受体（IGF1R）的潜在酪氨酸激酶双重抑制剂，这是一组在不同癌症（如胰腺导管腺癌（PDAC））中过表达的蛋白质。
