Faculty of Mathematics and Computer Science, Jagiellonian University, Łojasiewicza 6, 30-348 Kraków, Poland.
Maj Institute of Pharmacology, Polish Academy of Sciences, Smętna 12, 31-343 Kraków, Poland.
J Chem Inf Model. 2024 Mar 11;64(5):1425-1432. doi: 10.1021/acs.jcim.3c02038. Epub 2024 Feb 19.
Great progress in the development of computational strategies for drug design applications has revolutionized the process of searching for new drugs. Although the focus of strategies is still put on the provision of the desired activity of a compound to the considered target, characterization of a compound in terms of its physicochemical and ADMET properties becomes an indispensable element of computer-aided drug design protocols. In the study, an online application ADMET-PrInt for assessment of selected compound features: cardiotoxicity, solubility, genotoxicity, membrane permeability, and plasma protein binding was prepared. In addition to the prediction of particular property, ADMET-PrInt enables also the identification of compound features influencing this property thanks to the application of two explainability approaches: local interpretabile model-agnostic explanations and counterfactual analysis. It is an important factor for medicinal chemists, as it greatly facilitates the process of optimization of the compound structure in terms of the evaluated properties. The intuitive webpage, available at admet.if-pan.krakow.pl, allows making use of all predictive and interpretability models also by nonexperts and nonprogrammers.
在药物设计应用的计算策略开发方面取得的巨大进展,彻底改变了寻找新药的过程。尽管策略的重点仍然放在提供目标所需的化合物的活性上,但化合物在物理化学和 ADMET 性质方面的特征描述已成为计算机辅助药物设计方案不可或缺的组成部分。在这项研究中,我们开发了一个在线应用程序 ADMET-PrInt,用于评估所选化合物的特征:心脏毒性、溶解度、遗传毒性、膜通透性和血浆蛋白结合。除了预测特定性质外,ADMET-PrInt 还可以通过应用两种可解释性方法:局部可解释的模型不可知解释和反事实分析,来识别影响该性质的化合物特征。这对药物化学家来说是一个重要因素,因为它大大简化了根据评估性质优化化合物结构的过程。直观的网页,可在 admet.if-pan.krakow.pl 上访问,允许非专家和非程序员也使用所有预测和可解释性模型。