基于机器学习的假尿苷位点预测的研究进展

Research Progress of Pseudouridine Site Prediction Based on Machine Learning

摘要: 在基因的转录过程中，RNA很容易发生修饰的现象。迄今为止，研究人员已经发现了一百多种RNA的修饰，而假尿苷(ψ)是第一个被发现的，并且是目前存在最广泛的一种RNA修饰。近年来，随着表观遗传学研究的深入，关于假尿苷的研究越来越多。假尿苷修饰对于各种细胞生物和生理过程是至关重要的，研究的关键步骤就是在转录组中准确地识别出假尿苷的位点。由于实验化学方法来识别假尿苷位点耗时耗力，基于机器学习的计算方法来识别假尿苷位点是如今最好的选择。本文回顾了基于机器学习的假尿苷位点预测的研究现状，调查了研究人员在位点预测过程中使用的数据集和评估方法，得到了假尿苷位点预测的最新进展。本文选取具有代表性的几个机器学习模型进行简要概述，并对目前的局限性给出一些建议。

Abstract: RNA is easily modified in the process of gene transcription. To date, researchers have found more than a hundred RNA modifications, and pseudouridine (ψ) was the first to be discovered and is the most widely available type of RNA modification. In recent years, with the development of epigenetics, more and more studies on pseudouridine have been conducted. Pseudouridine modification is essential for various cellular biological and physiological processes, and the key step of the study is to accurately identify pseudouridine sites in the transcriptome. Since it is time-consuming and labor-intensive to identify pseudouridine sites by experimental chemical methods, the computational method based on machine learning is the best choice to identify pseudouridine sites today. This paper reviews the research status of pseudouridine site prediction based on machine learning, investigates the data sets and evaluation methods used by researchers in the process of site prediction, and gets the latest progress on pseudouridine site prediction. In this paper, several representative machine learning models are selected to give a brief overview and provide some suggestions on the current limitations.