Developing new algorithms and concepts towards understanding protein folding, misfolding, and aggregation

蛋白质在完成生命所必需的任务中是必不可少的，尽管它们的正常功能只有在折叠成天然状态时才能发挥作用。虽然天然状态在生理条件下是稳定的，但环境因素，如pH值、盐浓度或与细胞膜的相互作用，可能导致蛋白质片段的空间重排，并随后重新折叠和聚集成非天然的、通常是不需要的结构。蛋白质折叠、错误折叠和聚集在广泛的领域和应用中起着至关重要的作用：朝着推进蛋白质工程（一个新兴的数十亿美元的产业）；减少影响生物制药开发各个阶段并与高成本相关的蛋白质聚集；进一步加深我们对蛋白质错误折叠疾病的理解，这种疾病折磨着50多万加拿大人。在过去的50年里，我们对驱动蛋白质行为的基本过程的理解发生了重大变化。然而，令人惊讶的是，在能够从不同的初始构象从头预测折叠（和错误折叠）过程的模型方面，几乎没有取得什么进展。最先进的模型需要大量的计算能力，目前仅限于短蛋白质链，因此不能捕获所有涉及的机制。我的实验室的目标是开发新的概念性方法来理解蛋白质折叠、错误折叠和聚集的机制，这将允许在合理的时间内使用台式计算机对大型蛋白质进行计算机研究。从长远来看，这些方法将用于理解和预测更复杂的生物分子现象（如对接，蛋白质- dna相互作用，以及生物传感器应用中与纳米颗粒的相互作用）。在提议的程序中，我概述了使用各种计算工具和方法建模蛋白质行为的新方法。我的初步工作表明，如果在非笛卡尔坐标下进行模拟，蛋白质折叠预测所需的计算时间可以显着减少。将这项工作与数据挖掘工具相结合，我最近揭示了螺旋蛋白结构与其形成恶性错误折叠聚集体的倾向之间的关系。在此研究的基础上，该计划将成为3篇博士论文的基础。从长远来看，学生将在分子模型、数据挖掘和机器学习的发展方面接受培训。数据挖掘算法将用于识别蛋白质数据库中的模式，以揭示在环境压力下容易发生错误折叠和聚集的序列。机器学习算法将用于识别可靠的折叠路径。这项研究将促进对蛋白质行为的基本理解，同时培养工程学学生重要的计算技能。

Proteins are essential in performing tasks that are necessary for life, though their proper function can only be performed when folded into their native state. While the native state is stable under physiological conditions, environmental factors such as pH, salt concentration, or interaction with cell membranes may lead to spatial rearrangement of segments within the protein and its subsequent refolding and aggregation into nonnative and often undesirable structures. Protein folding, misfolding, and aggregation plays a critical role in a wide spectrum of fields and applications: Towards advancing protein engineering (a rising multibillion dollar industry); in diminishing protein aggregation that impacts biopharmaceutical development in every stage and is linked to high costs; and in furthering our understanding of protein misfolding disease that afflict over half a million Canadians. Our understanding of the fundamental processes that drive protein behavior has evolved significantly over the past 50 years. Nonetheless, surprisingly little advancements have been made in models that can predict the folding (and misfolding) process de novo, from a different initial conformation. State-of-the art models require substantial computational power and are currently limited to short protein chains, and hence cannot capture all the mechanisms involved. The goal of my lab is to develop new conceptual approaches to understand the mechanisms of protein folding, misfolding, and aggregation that will allow in silico studies of large proteins within a reasonable timeframe using a desktop computer. In the longer run, these methods will be used to understand and predict more complex biomolecular phenomena (such as docking, protein-DNA interactions, and interactions with nanoparticles for biosensor applications). In the proposed program I outline novel approaches for modeling protein behavior using various computational tools and methods. My preliminary work shows that computation time required for protein folding prediction can be significantly reduced if simulated in non-Cartesian coordinates. Combining this work with data mining tools, I recently revealed a relation between helical protein structures and their tendency to form malignant misfolded aggregates. Building on this research, the proposed program will form the basis of 3 PhD theses. The students will be trained in the development of molecular models, data mining, and machine learning in the longer run. Data mining algorithms will be used to recognize patterns within protein databases to reveal sequences that are liable to misfolding and aggregation under environmental stress. Machine learning algorithms will be used to identify reliable folding pathways. This research will advance basic understanding of protein behavior while fostering important computational skills in engineering students.