Designing AI-generated antimicrobials for targeting bacterial microdomains

设计人工智能生成的抗菌药物以靶向细菌微区

阅读:1

Abstract

Bacterial membranes are essential for processes such as nutrient transport, signal transduction, and maintaining cellular integrity. Within these membranes, specialized microdomains like cardiolipin-rich domains, serve as key sites for protein localization, membrane curvature regulation, and stress response. These microdomains possess high anionic headgroup density and specific lipid composition, which can enhance electrostatic and hydrophobic interactions with antimicrobial agents. In this study, we investigated whether antimicrobial compounds can selectively target membranes with distinct lateral lipid distributions. To this end, we modeled two bacterial membrane systems: one with a randomized lipid composition and another featuring an idealized cardiolipin-rich microdomain. Using a generative neural network framework guided by specific molecular design criteria, we generated AI-driven antimicrobial candidates and assessed their membrane interactions via free-energy calculations. Our analysis revealed preferential binding of compounds to cardiolipin-rich domains, evidenced by lower binding energies, alongside higher translocation barriers in these regions, attributable to strong electrostatic anchoring and tight lipid packing. Further clustering and feature importance analysis identified recurring structural motifs associated with potent antimicrobial activity, supporting that cardiolipin-rich regions may facilitate selective binding. Parallel toxicity predictions indicated several top candidates have low predicted toxicity, a critical factor for drug development. Notably, the top AI-designed compounds were benchmarked against established membrane-targeting drugs to assess similarities in their antimicrobial activity profiles. By integrating generative AI with advanced membrane modeling, this work establishes a robust framework for the rational design of new membrane-targeting antimicrobials and highlights how membrane composition influences drug-membrane interactions and efficacy.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。