Protein/lipid interactions: Determinants of lipid interactions with membrane proteins investigated by machine learning, molecular simulations and mass

蛋白质/脂质相互作用：通过机器学习、分子模拟和质量研究脂质与膜蛋白相互作用的决定因素

• 批准号: 2271160
• 金额: --
• 依托单位: University of Leeds
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2271160
• 关键词: Protein; lipid; interactions; Determinants; membrane

Background: Despite a fast-growing number of data that demonstrate that interactions of membrane proteins with lipids regulate their function (e.g. cholesterol or PIPs regulates the function of signaling receptors), the molecular/chemical details of such interactions remain elusive. Molecular dynamics simulations (MDS) have become an established technique for predicting protein/lipid interactions but they are too computationally prohibitive. Additionally, it is often challenging to use lab-based methodologies to study protein/lipid interactions. These are major limitations that impede the research.Objectives: Our aim is to combine MDS and mass spectrometry with artificial intelligence (AI)/machine learning (ML) to create a new approach that will significantly accelerate the prediction processes for protein/lipid interactions. In this approach we will use MDS to identify structural motifs on proteins that interreact with specific lipids using a set of proteins for which their 3D structure is known; the AI methods can learn from such data and predict lipid binding sites for other similar proteins.Novelty and Timeliness:Given the large increase of 3D membrane protein structures (some in complex with lipids), this research is timely in utilizing the cutting-edge AI/ML technologies to identify structural motifs on membrane proteins that interact with specific lipid types. Experimental approach: The student will use known 3D protein structures from the PDB and molecular simulations to identify how regions of different membrane protein families interact with specific lipid types. Then, AI/ML approaches will be developed to learn the interactions, to identify patterns in protein/lipid interactions, and to provide predictions for the interactions of other proteins. Native mass spectrometry will be used to evaluate and refine some of the results of the AI/ML methodology.

背景：尽管越来越多的数据表明膜蛋白与脂质的相互作用调节它们的功能(例如胆固醇或PIP调节信号受体的功能)，但这种相互作用的分子/化学细节仍然不清楚。分子动力学模拟(MDS)已成为预测蛋白质/脂类相互作用的成熟技术，但它们的计算量太大。此外，使用以实验室为基础的方法来研究蛋白质/脂类相互作用通常是具有挑战性的。这些都是阻碍这项研究的主要限制。目的：我们的目标是将MDS和质谱学与人工智能(AI)/机器学习(ML)相结合，创建一种新的方法，大大加快蛋白质/脂类相互作用的预测过程。在这种方法中，我们将使用MDS来识别蛋白质上与特定脂类相互作用的结构基序，使用一组已知其3D结构的蛋白质；AI方法可以从这样的数据中学习并预测其他类似蛋白质的脂类结合部位。新颖性和及时性：鉴于3D膜蛋白结构的大量增加(有些与脂类复杂)，本研究适时地利用尖端AI/ML技术来识别与特定脂类相互作用的膜蛋白上的结构基序。实验方法：学生将使用PDB中已知的3D蛋白质结构和分子模拟来确定不同膜蛋白家族的区域如何与特定的脂质类型相互作用。然后，将开发AI/ML方法来学习相互作用，识别蛋白质/脂类相互作用的模式，并为其他蛋白质的相互作用提供预测。原生质谱学将用于评估和提炼AI/ML方法的一些结果。