Computational modulator design and machine learning to target protein-protein interactions

摘要 我的研究计划的总体目标是开发和应用最先进的机器学习， 分子建模工具，以促进重要细胞途径的调节剂的合理设计， 治疗用途。蛋白质-蛋白质相互作用（PPI）是细胞信号传导和生物学过程中的核心因素。 网络，以及他们的错误管理导致疾病状态。因此，PPI是生物学上令人信服的靶点， 药物发现。尽管取得了一些显著的成功，但大多数生产者价格指数没有成功地成为目标， 治疗干预的挑战。最根本的挑战来自其内在的结构性 特点：许多PPI的结合面一般面积大、平整、动态。PPI通常 短暂的并且涉及多价接触。最有希望的PPI抑制剂发现策略之一是 使用微型蛋白质结构域模拟物（PDM）来复制自然界使用的关键界面接触。 PDM作为具有高表面互补性和更广泛的表面互补性的中等大小的分子是有利的。 接触点，但仍然是有限的，因为-根据定义-只有一部分的 总PPI结合能在相互作用中被捕获。合成结构域的结合亲和力通常是 低于同源全长蛋白质。在过去的五年里，我们大大推进了一个口袋引导 基于AlphaSpace的合理设计方法来应对这一挑战。我们成功地优化了 PDM通过引入非天然氨基酸来靶向p300/CBP共激活因子的KIX结构域，以改善 口袋片段结合;合理设计一种新型NEMO卷曲螺旋模拟物，破坏病毒诱导的NF-κB 并成功靶向MDM 2和MDMX上的新结合口袋， 有效的双重抑制剂，通过阐述氢键稳定的α-螺旋模拟物。与此同时， 开发了最先进的蛋白质配体对接以及虚拟筛选的评分功能，先进的 深度学习模型来预测分子性质和化学反应，并建立了强大的 与化学生物学和生物物理学的几个优秀实验室进行了富有成效的合作， 生物分子相互作用的新调节剂。这些重大进展为我们进一步推动 这是将机器学习和分子建模结合起来进行合理药物设计的前沿。我们专注于 未来几年将建立一个强大的基于AlphaSpace和机器的口袋引导设计平台 学习PPI正构抑制剂优化，提供物理/化学见解并开发新的 用于变构调节剂发现的计算策略，并探索具有深度的化学空间 多目标分子设计的序列/图形/几何表示学习。调制器设计 与我们的实验同事密切合作的努力不仅将严格测试预测能力， 我们开发的方法在真实的生活中的应用，而且还导致高度特异性和有效的调节剂， 几个重要但具有挑战性的治疗靶点，为药物开发提供了新的线索。

Abstract The overall goal of my research program is to develop and apply state-of-the-art machine learning and molecular modeling tools to facilitate the rational design of modulators of important cellular pathways for therapeutic use. Protein-protein interactions (PPIs) are central factors in cellular signaling and biological networks, and their mis-regulations lead to diseases states. Thus PPIs are biologically compelling targets for drug discovery. Despite a few notable successes, most PPIs have not been successfully targeted and remain challenging for therapeutic intervention. The fundamental challenge derives from their intrinsic structural features: the binding surfaces of many PPIs are generally large in area, flat, and dynamic. PPIs are often transient and involve multivalent contacts. One of the most promising PPI inhibitor discovery strategies is to use miniature protein domain mimetics (PDMs) to reproduce the key interface contacts utilized by nature. PDMs are advantageous as medium-sized molecules with high surface complementarity and a broader set of contact points than typical small molecules, but are still limited because—by definition—only a portion of the total PPI binding energy is captured in the interaction. The binding affinity of the synthetic domains is often lower than the cognate full-length proteins. In last five years, we have significantly advanced a pocket-guided rational design approach based on AlphaSpace to tackle this challenge. We have successfully optimized a PDM to target the KIX domain of the p300/CBP coactivator by introducing non-natural amino acids to improve pocket-fragment binding; rationally designed a novel NEMO coiled coil mimic that disrupts virus-induced NF-κB signaling and induces cell death; and successfully targeted a new binding pocket on MDM2 and MDMX with a potent dual inhibitor by elaborating hydrogen-bond stabilized alpha-helix mimetics. Meanwhile, we have developed state-of-the-art scoring functions for protein-ligand docking as well as virtual screening, advanced deep learning models to predict molecular properties and chemical reactions, and established strong and fruitful collaborations with several outstanding experimental labs in chemical biology and biophysics to discover new modulators of biomolecular interactions. These significant advances set the stage for us to further push the frontier of integrating machine learning and molecular modeling for rational drug design. Our focus in the next few years will be to establish a robust pocket-guided design platform based on AlphaSpace and machine learning for PPI orthosteric inhibitor optimization, provide physical/chemical insights and develop novel computational strategies for allosteric modulator discovery, and explore chemical space with deep sequence/graph/geometric representation learning for multi-objective molecular design. Our modulator design efforts in close collaborations with our experimental colleagues will not only rigorously test predictive power of our developed methods in real life applications, but also result in highly specific and potent modulators towards several important but challenging therapeutic targets, providing new leads for drug development.