基于深度学习的蛋白质翻译后修饰识别方法研究

Post translational modification is one of the most significant issues in the field of bioinformatics and computational biology. However, some limitations, which include low-accuracy identification of modification sites, modeling and analyzing the functions of post translational modification, can hardly be ignored. This work will focus on the modelling and analyzing post translational modification sites. The goals of this work are aimed at the effective identification and function discovery in the level of post translational modification. Firstly, we will research the local features of protein sequences and find out the potential key features with the deep learning approaches to finding out several significant and potential information in the level of protein sequences. And then, we will find out the differences and the commons of the various types of modification among different species and we try to discover their common features and specifically features. Last but not least, we try to model the relationships among the post translational modification types and protein functions in the level of multi-scale features. With the above mentioned project, several potential features could be discovered and mined in this field. Such plan can make contributions to understanding the protein function in the level of post translation modification. What’s more, it may be helpful to the development of precision medical and drug discovery.

蛋白质翻译后修饰的相关研究是生物信息学、计算生物学研究的热点问题之一。针对蛋白质翻译后修饰位点精准识别、建模分析等问题是当前相关领域亟待解决的问题。本项目拟针对蛋白质翻译后修饰位点为研究对象，开展识别位点有效识别和功能分析的机理挖掘，主要包括：（1）研究蛋白质序列的局部特征信息，通过深度网络挖掘“潜在”关键信息，从而发现蛋白质序列的“暗”信息；（2）研究不同物种的多种蛋白质翻译后修饰位点的共性与差异，构建通用型识别模型，构建不同物种不同翻译后修饰类型间的差异关系；（3）研究多尺度信息背景下，蛋白质翻译后修饰与蛋白质功能的关系建模和分析。从而帮助生物学家以及医学领域相关专家更精确地理解和挖掘蛋白质的潜在机制，分析不同生理和疾病条件下的蛋白质的功能差异，为个性化医疗的发展和新药开发设计提供帮助。

蛋白质翻译后修饰的相关研究是生物信息学、计算生物学研究的热点问题之一。针对蛋白质翻译后修饰位点精准识别、建模分析等问题是当前相关领域亟待解决的问题。本项目针对多种生物的蛋白质翻译后修饰各种位点进行识别，通过在一个物种上构建出较为精准的识别模型从而迁移到其他物种上进行类似的模型构建，从而构建能够适合多物种的蛋白质翻译后修饰位点识别模型。在整合传统分类模型的基础上，结合多组学特征数据，设计出了多种分类模型集成的多类型蛋白质翻译后修饰位点识别模。该思想推广到生物大分子的序列识别，从而构建出基于深度模型的生物序列标志物挖掘模型，并在此基础上，构建出关于药物和蛋白质相关性网络草图，为实现疾病生物标志物发掘奠定基础。