Leveraging biomolecular simulations to understand and predict the Blood-Brain-Barrier permeability of drugs

研究背景血脑屏障（BBB）是内皮细胞的集合，它选择性地调节任何特定物质从血液流入中枢神经系统（CNS）。依赖于药物进入中枢神经系统的医学治疗的关键是药物穿过血脑屏障的能力。近年来，关于药物血脑屏障通透性的可获得数据有所增加。这导致了预测血脑屏障渗透率的机器学习（ML）方法的新研究。然而，目前预测血脑屏障对给定药物的通透性的ML方法似乎在准确性和缺乏可解释性方面已经达到了极限。该项目的目的将是双重的，既提高血脑屏障渗透率的ML模型的准确性和可解释性。我们最终寻求理解的结构-功能关系的核心，在一个给定的药物（能力）通过血脑屏障。实现这一目标的主要方法是将机器学习与脑屏障的生物分子模拟相结合。通过利用华威大学的高性能计算资源，该项目希望在增强的采样技术的帮助下，全面模拟数百种药物分子的动力学。这比任何现有的技术都要大得多，因为目前的技术一次只能局限于少量的分子。该项目将对BBBoermeability动力学产生无与伦比的新见解，并彻底改变当前的ML建模范式。血脑屏障（BBB）是一层薄薄的细胞，将脑细胞与血液分离开来。为了有效地开发治疗脑部疾病的药物，这些药物需要能够穿过血脑屏障。近年来，随着数据的可用性提高了数据驱动技术来预测药物是否可以穿过血脑屏障。然而，这些模型似乎已经达到了极限性能。他们也不能解释为什么一个分子能或不能通过血脑屏障。本项目旨在利用BBB的大规模模拟以及新的数据驱动技术来开发BBB建模的新范式。这些技术在预测药物是否能通过血脑屏障方面应该更准确。同时还提供了对其预测背后原因的解释。研究方法的新颖性目前，药物-血脑屏障的模拟存在。但它们只能应用于一小部分药物分子。该项目旨在使用新的模拟方法和高性能计算来模拟数百种药物分子。这也可以用来为新的数据驱动方法提供信息，并开发更准确的方法。改进血脑屏障通透性的模型和理解可以大大降低药物开发成本并加快药物开发。这将引起阿斯利康（Astrazeneca）等制药公司的极大兴趣，Sosso博士集团正与阿斯利康积极合作。该研究属于EPSRC的生物化学和生物信息学以及计算化学的职权范围外部合作伙伴-阿斯利康作为制药领域的主要参与者之一，阿斯利康显然对提高机器学习在预测药物渗透血脑屏障能力方面的当前能力感兴趣。这个项目特别寻求超越目前的技术水平，利用大规模的分子动力学模拟来增强我们目前可用的数据集，并了解药物样分子在血脑屏障渗透的核心机制。因此，这个项目的成果为现实世界的影响奠定了一条非常具体的道路。

The context of the researchThe Blood-Brain-Barrier (BBB) is a collection of endothelial cells that selectively regulates the influx of any given substance from our blood into the central nervous system (CNS). Crucial to medical treatments relying on the delivery of medicinal drugs in to CNS is the capability of siad drugs to actually cross the BBB.In recent years, there has been an increase in pibicly available data on the BBB permeability of drugs. This has led to novel research in Machine Learning (ML) methods of predicting BBB permeability. However, current ML methods of predicting the permeability of the BBB to a given drug seem to have reached its limit in accuracy and lack explainability.The aim of the project will be twofold, to both improve the accuracy and explanability of ML models of BBB permeability. We ultimately seek to understand the structure-function relation at the heart of a given drug (in-ability) to cross the BBB. The primary method for this will ne through combining machine learning with biomolecular simulations of the BBB. Through leveraging the high-performance computational resources at Warwick, this project hopes to fully simulate dynamics for hundreds for drug molecules potentially aided by enhanced sampling techniques. This is far a greater number than any exsiting technique as present techniques are limited to a small handful of molecules at a time. This project will yield unparallel novel insight into kinetics of BBBoermeability, as well revolutionise the current ML moldelling paradigm.The blood brain barrier (BBB) is a thin layer of cells which seperates brain cells from the bloodstream. In order to effectively develop drugs whihc treat ailments of the brain, these drugs need to be able to cross the BBB. In the recent years, as availabilty of data has improved data driven techniques to predict whether a drug can cross the BBB. However, these models seemed to have reached a limit performance. They can also can't explain why a molecule can or cannot pass through the BBB.The aims and objectives of the researchThis project aims to use large scale simulations of the BBB along with new data driven techniques to develop new paradigms of BBB modelling. These techniques should allow better accuracy in predicting whather a drug can pass through the BBB. along with providing explanations as to the reasoning behind is predictions.The novelty of the research methodologyCurrently, sumulations of drugs-BBB exist. but they can only be applied to a small handful of drug molecules. This project aims to use new methods of simulation with high-performance computing to simulate hundreds of drug molecules. this can also be used to inform new data driven approaches and develop more accurate methods.The potential impact, applications and benefitsAn improved model and understanding of BBB permeability can greatly reduce the cost of and accelerate drug development. This would be of great interest to pharmacetical companies such as Astrazeneca, with whom Dr Sosso's Group is actively collaborating.How the research relates to the remitThe research falls into both the EPSRC remit of biological chemistry and biological informatics as well as computational chemistryExternal Partner - AtstraZenecaAs one of the major players in the context of Pharmceuticals, AstraZeneca has obvious interest in improving on the current capabilities of Machine Learning on terms of predicting the ability of drugs to permeate tje Blood-Brain-Barrier. This project specifically seeks to go beyond the state of the art leveraging large scale molecular dynamic simulations to both enhance the datasets available to us at the moment and to understand tje mechanism(s) at the heart of the Blood-Brain-Barrier permeation by drug-like molecules. thus, the outcomes of this project layout a very concrete path toward real-world impact.