Machine Learning Guided Biophysical Model Development of Amino Acid and tRNA Effects on Translation-Elongation Speed

该项目利用人工智能来加速发现核糖体（一种复杂的分子机器）控制蛋白质合成的因素。可以快速生成关于细胞的分子生物学的大量数据。然而，理解大量数据以获得理解是一个重大挑战。具体来说，在核糖体合成蛋白质的过程中，有如此多的分子因子相互作用，以至于很难确定哪些分子因子调节核糖体的功能速度。在这个项目中，人工智能被用来识别假定的因果特征，然后成为开发基于物理和化学的模型的起点，这些模型可以解释这些变量之间的物理关系和核糖体功能的速率。 由于这种方法是通用的，它将在主题之间转移，从而加速从数据到洞察一系列问题的过程。这项研究将使预测氨基酸突变对蛋白质合成的影响成为可能。生物工程和生物制药界可以利用这些信息来优化蛋白质表达。最后，该提案将通过教授来自代表性不足的群体的高中生机器学习主题并使他们对STEM领域感兴趣来促进科学的多样性。基于化学和物理的生物分子过程模型是整个生物化学和分子生物学社区用于解释分子行为和实验数据之间关系的关键工具。发展这种模型的一个瓶颈是识别驱动感兴趣的生物分子过程的基本特征。机器学习模型通常可以做出准确的预测，但没有解释能力，因此有可能快速识别这些基本特征。该项目将创建一个工作流程，利用机器学习的这一有益功能来指导生物物理模型的开发，从而加速从数据到洞察力的过程。PI的实验室最近证明，核糖体的A位点和P位点中的转移RNA和氨基酸的同一性可预测地和因果地调节A位点的翻译延伸速度。该项目将应用机器学习工作流程来建模和理解这种效应的分子起源，这是目前未知的。首先，将利用集成机器学习方法，识别氨基酸和tRNA分子在E-、P-和A-位点的稳健的物理化学特征，从而准确预测翻译速度。接下来，这些强大的和预测的物理化学性质将被用作构建物理模型的起点，解释为什么这些性质是重要的。最后，来自模型的预测和见解将由合作者在体内进行实验测试。该奖项反映了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。

This project leverages artificial intelligence to accelerate the discovery of factors governing protein synthesis by the ribosome, a complex molecular machine. Large amounts of data can be rapidly generated concerning the molecular biology of the cell. Yet making sense of massive amounts of data to gain understanding is a significant challenge. Specifically, during the process of protein synthesis by the ribosome there are so many molecular factors interacting that determining which of those regulate the speed at which it functions is difficult. In this project, artificial intelligence is used to identify putative causal features, which then become the starting point for the development of physics and chemistry based models that can explain the physical relationship between those variables and the rate at which the ribosome functions. Because this approach is general, it will be transferable between topics, thereby accelerating the process of going from data to insight across a range of problems. This research will make it possible to predict the influence of amino acid mutations on protein synthesis. Bioengineering and biopharmaceutical communities can exploit this information for optimization of protein expression. Finally, this proposal will promote diversity in the sciences by teaching high-school students from underrepresented groups topics in machine learning and interest them in STEM fields.Chemistry- and physics-based models of biomolecular processes are critical tools used throughout the biochemistry and molecular biology communities to explain the relationship between molecular behaviors and experimental data. A bottleneck in the development of such models is the identification of the essential features driving the biomolecular process of interest. Machine learning models, which often make accurate predictions with no explanatory power, offer the potential to rapidly identify these essential features. This project will create a workflow that will leverage this beneficial feature of machine learning to guide biophysical model development, and thereby accelerate the process of going from data to insight. The PI’s lab recently demonstrated that the identity of the transfer RNAs and amino acids in the A- and P-sites of the ribosome predictably and causally modulate the translation elongation speed at the A-site. This project will apply the machine learning workflow to model and understand the molecular origins of this effect, which are currently unknown. First, an ensemble machine learning approach will be utilized that identifies the robust physicochemical features of amino-acid and tRNA molecules at the E-, P- and A-sites that accurately predict translation speed. Next, these robust and predictive physicochemical properties will be used as a starting point to construct physical models that explain why those properties are important. Finally, the predictions and insights from the models will be experimentally tested in vivo by a collaborator.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.