NOT-GM-21-028:Towards a generalizable drug discovery framework based on intrinsically disordered regions

项目摘要 目前的药物发现方法产生的收益递减，因为成本，失败率， 耐药性都在增加。与此同时，新的靶点和候选药物没有跟上 满足了所有疾病的需求。这项工作旨在解决其中几个领域的问题。 它旨在降低成本，提高成功率，并解决耐药性，同时增加 新的靶点和潜在的候选药物。 虽然大多数药物是通过试错法发现的，或者是针对特定结构的蛋白质口袋设计的， 事实证明，许多疾病和耐药性发生在涉及无序的界面， 蛋白质区域。因此，虽然大多数药物设计的信息学都集中在结构蛋白质上， 口袋，一个具有巨大潜力的领域在于无序蛋白质及其界面。到 要有效地、大规模地这样做，就需要一个信息学框架， 跨基因组、蛋白质组、结构、化学、途径、本体论、相互作用的信息 建模和进化空间。 在这里，我们提出了这样一个框架，创建：1）无序的目标库， 2）可以模拟的小分子， 这些蛋白质可以与无序区域结合，从而与无序蛋白质区域的通常伴侣相互作用。 我们将首先在几种生物体中创建一个无序的目标库。然后，通过一个 贝叶斯框架，我们将整合专家知识，序列信息/统计， 相互作用模型来预测药物，这些药物可以：1）靶向这些区域，2）模拟这些区域， 区域互动。最后，我们将重点关注耐药病原体，以验证预测 药物实验

Project Summary Current approaches to drug discovery are yielding diminishing returns as costs, failure rate, and drug resistance all increase. Meanwhile, novel targets and drug candidates are not keeping up with demand across the disease spectrum. This work seeks to address several of these areas. It seeks to lower cost, increase success rate, and address drug resistance while increasing novel targets and potential drug candidates. While most drugs are found by trial-and-error or designed for specific structured protein pockets, it turns out that many diseases and drug resistance occur at interfaces involving disordered protein regions. So, while most informatics for drug design has focused on structured protein pockets, an area with tremendous potential lies in disordered proteins and their interfaces. To do so effectively, and at a large-scale, an informatics framework is needed that effectively uses information across genomic, proteomic, structural, chemical, pathway, ontological, interaction modeling, and evolutionary space. Here, we present such a framework that creates: 1) disordered target libraries and corresponding small molecules to interact them and 2) small molecules that can mimic disordered regions and thus interact with the usual partners of the disordered protein regions. We will first create a disordered target library across several organisms. Then, through a Bayesian framework, we will integrate expert knowledge, sequence information/statistics, and interaction modeling to predict drugs that can: 1) target these regions and 2) mimic these regions in interactions. Finally, we will focus on drug resistant pathogens to validate predicted drugs experimentally.