基于RNA-Seq数据的关键选择性剪接识别方法研究

发布时间：2018-11-04 10:31

【摘要】：选择性剪接是指在基因表达过程中,同一段基因编码区域能够经转录并翻译成多个具有不同功能的蛋白质的一种细胞内在机制。识别选择性剪接事件对研究基因功能、蛋白质结构多样性、细胞分化、物种进化具有重要意义。随着高通量测序技术的问世及快速发展,从转录组高通量测序(RNA-Sequencing,RNA-Seq)数据中识别选择性剪接事件成为了当前生物信息学研究的一个前沿课题。然而,如何准确地从RNA-Seq数据中识别外显子跳跃事件和内含子保留事件仍是一个未被解决的问题。现有的方法在识别外显子跳跃和内含子保留这两种选择性剪接事件上依然存在着诸多问题,例如:利用了与外显子跳跃事件或内含子保留事件相关的部分信息去构建计算模型和方法;使用了低质量的测序短片段;对特征的正则化表述方式没有统一的标准;没有指出哪一种特征对准确地识别选择性剪接事件是最有效的等。针对现有的基于RNA-Seq数据识别外显子跳跃事件和内含子保留事件的方法中所存在的问题,本文进行了深入的讨论与研究。本文的主要工作和创新点包括:(1)提出了外显子跳跃事件相关特征分析方法该方法的创新点包括:使用了外显子跳跃事件相关的多种特征来表述每一个外显子,评估了每一种特征对准确识别外显子跳跃事件的影响,同时构建4种特征集合来评估不同的特征正则化方法对准确识别外显子跳跃事件的影响。在真实的人类骨骼肌、大脑、心脏和肝脏组织的RNA-Seq数据中,整合现有方法的识别结果构建了参考集,本文评估了外显子跳跃事件相关特征在准确识别外显子跳跃事件中的重要性,并分析了测序短片段特征的不同正则化表述方式对准确识别外显子跳跃事件的影响。实验结果表明映射到支持外显子跳跃区域的、连接选择性外显子的上游外显子和下游外显子的测序短片段特征,以及该选择性外显子保留水平psi得分这两种特征对准确识别外显子跳跃事件具有重要影响。并且得出使用原始测序短片段数或是使用经过正则化的测序短片段数来描述特征对准确识别外显子跳跃事件没有显著影响。(2)提出了基于多特征分析的外显子跳跃事件识别方法EScallEScall方法的创新点包括:过滤掉映射质量数低的和具有歧义的测序短片段的映射结果,同时整合多种与外显子跳跃事件相关的特征,包括:映射到外显子内部区域的测序短片段特征、支持连接两个外显子的测序短片段特征、和基因表达信息等特征,设计了新的外显子跳跃事件的得分计算方法,用于从两种不同条件下的RNA-Seq数据中识别外显子跳跃事件。在真实的人类骨骼肌和大脑组织的RNA-Seq数据中,使用EScall方法识别外显子跳跃事件。将识别结果与其他方法的识别结果相比较,实验结果表明EScall方法能够有效地减少假阳性和假阴性的识别结果,获得了较好的预测结果。(3)提出了基于联合得分的内含子保留事件识别方法IRcallIRcall方法的创新点包括:整合7种与内含子保留事件相关的特征,包括:映射到内含子内部区域的测序短片段特征、支持内含子剪接的连接该内含子上游外显子和该内含子下游外显子的测序短片段特征、映射到该内含子上游和下游外显子内部区域的测序短片段特征、与5’端剪接位点相交叠的测序短片段特征、与3’端剪接位点相交叠的测序短片段特征、内含子区域能被测序短片段覆盖的比例特征、和基因表达信息特征,设计了新的内含子保留事件联合得分IRScore的计算方法,用于从两种不同条件下的RNA-Seq数据中识别内含子保留事件。在真实的拟南芥skip突变体和野生型的RNA-Seq数据中,使用IRcall方法识别内含子保留事件。将识别结果与其他方法的识别结果相比较,实验结果表明IRcall方法能够有效地减少假阳性的识别结果,从而更加准确地识别内含子保留事件。(4)提出了基于随机森林的内含子保留事件识别方法IRclassifierIRclassifier方法的创新点包括:整合3种方法的识别结果构建了参考集,使用与内含子保留事件相关的21种特征来表述每一个内含子,构建了基于随机森林的分类器,用于从两种不同条件下的RNA-Seq数据中识别内含子保留事件,同时分析了每一种特征对准确识别内含子保留事件的影响。在真实的拟南芥skip突变体和野生型的RNA-Seq数据中,整合现有三个识别方法在1号、2号、4号染色体上的识别结果来构建训练集。使用IRclassifier识别内含子保留事件,实验结果表明该方法识别结果的准确率达到99.2%。此外,使用IRclassifier识别3号和5号染色体上的内含子保留事件,将识别结果与现有方法的识别结果相比较,实验结果表明IRclassifier方法能够准确地识别内含子保留事件,验证了方法的有效性。
[Abstract]:Alternative splicing refers to a cellular intrinsic mechanism that can be transcribed and translated into a plurality of proteins having different functions during gene expression. Identifying alternative splicing events is of great significance to the study of protein structure, protein structure diversity, cell differentiation and species evolution. With the advent of high-throughput sequencing technology and rapid development, the identification of selective splice events from transcriptome high-throughput sequencing (RNA-Sequence, RNA-Seq) data has become a leading issue in the study of bioinformatics. However, how to accurately identify exon skipping events and introns retention events from RNA-Seq data remains an unresolved issue. The existing methods still exist many problems in identifying exon skipping and intron retention, for example, using partial information related to exon skipping events or intron retention events to build a computing model and method; A low-quality sequencing video segment is used; there is no uniform standard for the regularization of features; it is not pointed out which feature is the most effective for accurately identifying alternative splicing events. In view of the existing problems in the methods of identifying exon skipping events and introns retention events based on RNA-Seq data, this paper makes an in-depth discussion and study. The main work and innovation points of this paper include: (1) the relevant characteristic analysis methods of exon skipping events are put forward, and the innovation points of the method include: expressing each exon by using a plurality of characteristics related to exon skipping events, The effect of each feature on accurate identification of exon skipping events was assessed, and four feature sets were constructed to assess the effect of different feature regularization methods on accurately identifying exon skipping events. In the real human skeletal muscle, the RNA-Seq data of the brain, the heart and the liver tissue, the recognition results of the existing method are integrated into the reference set, In this paper, the influence of different regularized expressions of the characteristics of the short segment of sequencing on the jumping events of exons can be accurately identified. The results of the experiment indicate that the characteristics of the short segment of the upstream and downstream exons of the selective exon are mapped to the jump region of the support exon, and the two features of the selective exon reservation horizontal psi score have an important influence on the accurate identification of the exon skipping events. and it is concluded that the feature has no significant effect on the accurate identification of exon jump events using the number of original sequencing clips or using a normalized number of sequenced short segments. (2) An innovative point of EscoalEscoall method based on multi-characteristic analysis of exon skipping events is proposed, including: filtering out mapping results with low mapped mass numbers and ambiguous sequencing video segments, and combining a plurality of features related to exon skipping events, including: The method comprises the following steps of: mapping the characteristic of the sequencing short film segment to the inner region of the exon, supporting the characteristics of the sequencing film segment connecting the two exons, and the gene expression information and the like, and designing a score calculation method of the new exon jump event, for identifying exon skipping events from RNA-Seq data under two different conditions. In real human skeletal muscle and brain tissue RNA-Seq data, exon skipping events were identified using the Escoall approach. Comparing the recognition results with the recognition results of other methods, the experiment results show that the EScall method can effectively reduce false positive and false negative identification results, and obtain better prediction results. (3) A new method for the identification of intron retention events based on joint score is proposed, including: the feature of 7 species associated with intron retention events, including the feature of sequencing short segments mapped to introns interior regions, a sequencing video segment feature that supports intron splicing of the intron upstream exon and the intron downstream exon, a sequencing short segment feature mapped to an inner region of the intron upstream and downstream, a sequencing short segment feature that overlaps the 5 'end splice site, Compared with the 3' terminal splicing site, the sequence short segment feature, the intron region can be sequenced short segment coverage ratio feature, and the gene expression information characteristic, the calculation method of the new intron retention event joint score IRSore is designed, An intron retention event is identified from RNA-Seq data under two different conditions. In real Arabidopsis thaliana skip mutants and wild-type RNA-Seq data, introns retention events were identified using the IRXRF method. Comparing the recognition results with the recognition results of other methods, the experimental results show that the IR032 method can effectively reduce false positive identification results, thus more accurately recognizing intron retention events. (4) An innovative point of IRclassfier IRclassfier method based on an intron retention event recognition method based on a random forest is proposed, which comprises the following steps: a reference set is constructed based on the recognition results of the three methods, 21 features related to the intron retention event are used to express each intron, A random forest-based classifier was constructed for identifying intron retention events from RNA-Seq data under two different conditions while analyzing the effect of each feature on accurately identifying intron retention events. In the real Arabidopsis thaliana skip mutant and wild-type RNA-Seq data, the training set was constructed by combining the recognition results of the existing three identification methods on chromosomes 1, 2 and 4. Intron retention events were identified using IRclassfier, and the results showed that the accuracy of the method was 99.2%. In addition, using IRclassfier to identify intron retention events on chromosome 3 and chromosome 5, the results of the identification are compared with the recognition results of the existing methods, and the results show that the IRclassfier method can accurately identify intron retention events and verify the effectiveness of the method.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：Q78

【相似文献】