Leveraging evolutionary analyses and machine learning to discover multiscale molecular features associated with antibiotic resistance

总结 抗生素耐药性（AR）是一个高度优先的紧迫威胁。AR病原体，如ESKAPE组， 数百万人感染，数十万人死亡。虽然目前的战略，如遗传和药物 筛选有助于识别特定病原体中对AR至关重要的基因和突变，但广泛缺乏 帮助理解AR的起源和持续适应的方法。AR可以通过多种途径在病原体中产生。 分子变化，包括获得蛋白质结构域、单个基因或代谢能力。因此，我们认为， 预测和克服新出现的病原体中的AR或发现新的AR机制需要一个全面的 了解AR在多个分子尺度上的进化。然而，利用这些不同的数据集， 因为原始数据库彼此孤立。此外，不同的数据类型很难 以一种在生物学上有意义的方式跨尺度整合。在这个项目中，我们描述了一个计算发现， 结合进化分析和机器学习的框架，以跨多个尺度整合AR数据 获得对ESKAPE病原体中AR分子特征的机制见解，并在新的 （重新）出现的基因组。我们将把我们的方法作为开放的FAIR数据库、开放的软件， 计算、实验和临床AR社区的网络平台。我们将与AR密切合作 合作者、最终用户和开放软件社区，以确保 发布可访问的、用户友好的交互式平台。最后，在获奖后的扩展阶段，我们 将与NIAID资助的生物信息学联盟合作，进行下游数据和方法的整合， 长期可持续性。该框架将在本项目中开发，将广泛适用于推进 了解研究不足和新兴病原体（ESKAPE以外）中的AR，以结束武器 微生物和药物之间的竞争，创造更好的治疗效果。

Summary Antibiotic resistance (AR) is a high-priority urgent threat. AR pathogens such as the ESKAPE group cause millions of infections and hundreds of thousands of deaths. While current strategies such as genetic and drug screens have helped identify genes and mutations critical for AR in specific pathogens, there is a broad lack of methods to help understand AR’s origin and continuous adaptation. AR can arise in a pathogen via a variety of molecular changes, including acquiring protein domains, individual genes, or metabolic capabilities. Hence, predicting and overcoming AR in emerging pathogens or discovering new AR mechanisms requires a holistic understanding of AR evolution across multiple molecular scales. However, leveraging these diverse datasets is challenging because original databases are siloed from each other. Further, the different data types are hard to integrate in a biologically-meaningful way across scales. In this project, we describe a computational discovery framework combining evolutionary analyses and machine learning to integrate AR data across multiple scales to gain mechanistic insights into AR molecular features in ESKAPE pathogens and predict AR in new (re)emerging genomes. We will implement our approach as open FAIR data repositories, open software, and web platforms for the computational, experimental, and clinical AR communities. We will work closely with AR collaborators, end-users, and the open software community during and following the project duration to ensure the release of accessible, user-friendly, interactive platforms. Finally, in the post-award expansion phase, we will work with NIAID-funded bioinformatics consortia for downstream integration of data and methods and long-term sustainability. The framework will develop in this project will be broadly applicable to advance understanding of AR in understudied and emerging pathogens (beyond ESKAPE) towards ending the arms race between microbes and drugs by creating better treatment outcomes.