CAREER: Chemical Theory for the Protein Crystal Folding Problem

爱荷华大学的Michael Schnieders教授获得了化学系化学理论、模型和计算方法项目的奖励，以开发预测晶体结构的新理论方法。有机分子晶体在化学、生物化学、材料科学、药理学和工程等领域发挥着重要作用。有机分子晶体的一个日常例子是药物片剂，其典型配方是为了优化诸如保质期（即热稳定性）和溶解度（即摄入后溶解）等特性。有机晶体在通过x射线晶体学实验了解生物分子（即蛋白质）的结构和功能方面的关键作用可能不太为人所知。药物分子通常只由几十个原子组成，而蛋白质通常由数千个原子组成，这些原子的排列（即三维排列）用一种称为“蛋白质折叠”的过程来描述。蛋白质折叠背后的驱动力是疏水效应，这也是通常观察到的油和水分离趋势的原因。Schnieder博士团队的工作重点是将所有有助于蛋白质折叠的力严格结合到有效的算法中，用于计算预测肽和蛋白质晶体结构（多态）。该方法将通常用于预测小分子晶体多态性的先进分子相互作用模型与描述蛋白质折叠所需的复杂分子动力学采样算法相结合。该项目的影响是将晶体结构预测（CSP）领域的边界从小有机分子（即数十个原子）扩展到包括肽和蛋白质（即数百或数千个原子）。施耐德博士的研究与三管齐下的教育推广战略完全结合，加强并进一步多样化基于仿真的工程与科学（SBE&amp；S）的培训。该项目的教育计划包括：向代表性不足的高中生推广，帮助他们制作计算科学竞赛项目；开设现代计算生物化学课程，培训（本科生）毕业生应用SBE&amp；S方法解决计算（生物）化学的基本问题。第三个目标是继续传播开源的Force Field X软件（http://ffx.biochem.uiowa.edu）。在sbe&s和高性能计算（HPC）领域的领导地位对美国的全球竞争力至关重要。基于物理的蛋白质折叠通过分子动力学（MD）固有地解释了温度，压力，溶剂环境和熵的贡献，如疏水效应。另一方面，几乎所有当前的晶体结构预测（CSP）方法都是系统地或随机地搜索势能面，而不是自由能面，在有限的情况下，近似地包含熵考虑。本项目的前提是，对于“蛋白质晶体折叠问题”，普遍适用的解决方案需要在多晶发现模拟过程中有效地包含温度、压力和溶剂环境（疏水效应、pH等）。由于结晶成核动力学缓慢，普通无偏MD对CSP的效率不高。为了克服这个问题，一组新的算法正在开发中，以帮助打开聚合物晶体性质预测的大门。第一个目标集中在两个新的炼金术热力学路径，它不需要先验的结晶状态知识，并显著加速相变1)真空和结晶状态之间（即升华/沉积）和2)溶剂化和结晶状态之间（即溶解度）。两种途径都有效地包括温度和压力的影响，而后一种途径还包括溶剂环境的影响。第二个目标侧重于极化力场（例如变形虫）的第一个恒定pHMD （CpHMD）算法，以解释具有许多可滴定残基折叠和/或经历结晶相变的聚合物（例如蛋白质或核酸）的质子化变化。除了该项目对蛋白质晶体的关注之外，采样算法和CpHMD理论广泛适用于一系列模拟应用，包括蛋白质-配体结合，分子设计和针对实验的结构模型的改进（即x射线和中子晶体学，CryoEM， NMR等）。项目教育计划包括：1)向未被充分代表的高中学生推广，以促进计算科学博览会项目；2)创建现代计算生物化学课程，培养（本科生）毕业生应用sbe&&s方法解决计算（生物）化学的基本问题；3)继续传播开源的Force Field X软件（http://ffx.biochem.uiowa.edu）.This奖反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Professor Michael Schnieders of the University of Iowa is supported by an award from the Chemical Theory, Models, and Computational Methods program in the Division of Chemistry to develop new theoretical approaches to predict crystal structures. Organic molecular crystals play an important role in a range of fields including chemistry, biochemistry, materials science, pharmacology, and engineering. One everyday example of organic molecular crystals are pharmaceutical tablets, which are typically formulated to optimize properties such as shelf-life (i.e. thermal stability) and solubility (i.e. dissolution upon ingestion). A perhaps less appreciated role of organic crystals has been their pivotal impact in understanding the structure and function of biomolecules (i.e. proteins) via X-ray crystallography experiments. Whereas drug molecules typically consist of only a few dozen atoms, proteins generally consist of thousands of atoms whose packing (i.e. 3-dimensional arrangement) is described by a process called "protein folding". A driving force behind the folding of proteins is the hydrophobic effect, which is also responsible for the commonly observed tendency of oil and water to separate. The work in Dr. Schnieder's group focuses on the rigorous incorporation of all forces that contribute to protein folding into efficient algorithms for the computational prediction of peptide and protein crystal structures (polymorphs). The approach combines advanced models of molecular interactions commonly used to predict small molecule crystal polymorphs with sophisticated molecular dynamics sampling algorithms needed to describe protein folding. The impact of this project is to expand the boundaries of the crystal structure prediction (CSP) field beyond small organic molecules (i.e. dozens of atoms) to include peptides and proteins (i.e. hundreds or thousands of atoms). Dr. Schnieder's research is fully integrated with a three-pronged strategy for educational outreach that strengthens and further diversifies training in Simulation Based Engineering & Science (SBE&S). The project's educational plan includes: outreach to underrepresented high school students to help make computational science fair projects and creation of a modern Computational Biochemistry course to train (under)graduates in applying SBE&S methods to fundamental problems in computational (bio)chemistry. The third aim is the continued dissemination of open source Force Field X software (http://ffx.biochem.uiowa.edu). Leadership in SBE&S and high-performance computing (HPC) is of critical importance to the global competitiveness of the United States. Physics-based protein folding via molecular dynamics (MD) inherently accounts for temperature, pressure, solvent environment and entropic contributions such as the hydrophobic effect. On the other hand, nearly all current crystal structure predication (CSP) approaches perform either a systematic or stochastic search of a potential energy surface, rather than a free energy surface, followed in limited cases by approximate inclusion of entropic considerations. The premise of this project is that a generally applicable solution to the "protein crystal folding problem" requires efficient inclusion of temperature, pressure and solvent environment (hydrophobic effect, pH, etc.) during polymorph discovery simulations. Due to the slow nucleation kinetics of crystallization, ordinary unbiased MD is not efficient for CSP. To overcome this, a novel family of algorithms are being developed to help open the door to polymer crystal property prediction. The first objective focuses on two novel alchemical thermodynamic paths, which do not require a priori knowledge of the crystalline state and that dramatically accelerate phase transitions 1) between vacuum and crystalline states (i.e. sublimation/deposition) and 2) between solvated and crystalline states (i.e. solubility). Both paths efficiently include the influence of temperature and pressure, while the latter path additionally includes the influence of the solvent environment. The second objective focuses on the first constant pH MD (CpHMD) algorithms for a polarizable force field (e.g. AMOEBA) to account for protonation changes as a polymer (e.g. a protein or nucleic acid) with numerous titratable residues folds and/or undergoes a crystalline phase transition. Beyond the focus of this project on protein crystals, the sampling algorithms and CpHMD theories are broadly applicable to a range of simulation applications, including protein-ligand binding, molecular design and refinement of structural models against experiment (i.e. X-ray and neutron crystallography, CryoEM, NMR, etc). The project's educational plan includes: 1) outreach to underrepresented high school students to facilitate computational science fair projects, 2) creation of a modern Computational Biochemistry course to train (under)graduates in applying SBE&S methods to fundamental problems in computational (bio)chemistry, and 3) continued dissemination of the open source Force Field X software (http://ffx.biochem.uiowa.edu).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.