基于深度学习的长链非编码RNA与蛋白质相互作用关系预测的研究进展

Research Progress on Predicting Interaction between Long Noncoding RNA and Protein Based on Deep Learning

摘要: 基因表现型受调控基因的表达来驱动，而基因的表达同时也受不同生物分子之间复杂的相互作用所影响。21世纪以来，蛋白质-DNA相互作用以及RNA-RNA相互作用都收获了很好的研究成果。相比起来，在受制于各种困难的情况下，长链非编码RNA-蛋白质相互作用(LPI)机制相对未知。然而LPI正在成为遗传机制中的关键相互作用，在人类发育过程中扮演着越来越重要的角色，因此人们对于长链非编码RNA与蛋白质相互作用关系研究的热情不断高涨。由于使用大规模实验方法来验证两者的相互作用需要耗费很多物质成本和时间成本，与此同时伴随着当代社会科学技术不断发展，计算机硬件持续迭代更新，研究人员开始将重心放在如何使用机器学习及深度学习方法来进行基础预测，可喜的是这些预测实验都取得了令人满意的结果。为了丰富该领域的科研论文资料，方便科研人员了解深度学习技术在长链非编码RNA与蛋白质相互作用关系预测的应用，在此我们回顾了长链非编码RNA与蛋白质相互作用预测的研究现状，调查了近期研究人员在实验中利用到的最新方法和数据库，并总结了相关数据，提出了一些可行性建议。

Abstract: Gene phenotypes are driven by the expression of regulatory genes, which are also influenced by complex interactions between different biomolecules. Since the 21st century, protein-DNA interac-tion and RNA-RNA interaction have been well studied. In contrast, under many difficulties, the mechanism of long non-coding RNA-protein interaction (LPI) is relatively unknown. However, LPI is becoming a key interaction in the genetic mechanism and playing an increasingly important role in human development. Therefore, the research on the interaction between long non-coding RNAs and proteins is increasingly enthusiastic. Since the use of large-scale experimental methods to verify the interaction between the two requires a lot of material and time costs, at the same time, with the continuous development of contemporary social science and technology and the continuous iterative updating of computer hardware, researchers began to focus on how to use the machine learning and deep learning approach to predict. It is gratifying that these prediction experiments have achieved satisfactory results. In order to enrich scientific research papers in this field and facilitate researchers to understand the application of deep learning technology in predicting the interaction relationship between long non-coding RNA and protein, we reviewed the research status of long non-coding RNA and protein interaction prediction, investigated the latest methods and databases used by researchers in recent experiments, and summarized relevant data. Finally, some feasible suggestions are put forward.