中文摘要：

mRNA3’端的多聚腺苷酸化是真核细胞内mRNA转录后处理的三个最主要步骤之一。对DNA序列上发生多聚腺苷酸化的位置即PolyA位点的识别，对于理解mRNA的形成机制以及进行基因结构预测具有重要作用。本研究利用机器学习方法对PolyA位点进行预测，其实现过程分为以下三个步骤：特征的生成、特征的筛选、特征的综合分析聚类。首先，我们采取统计k阶核苷酸频率的方法来生成初始的特征；然后，通过信息学知识来对特征进行筛选；最后，使用SVM（Support Vector Machines，支持向量机）的方法进行特征的综合分析，确定参数，建立预测模型。在独立的测试数据集上进行测试，当敏感度（Sn）固定为60％时，在内含子水平和外显子水平上的特异性（sP）分别为71．67％和80．77％，在内含子水平上的预测精度明显优于国际上的同类软件。

英文摘要：

Polyadenylation （PolyA） occurs in mRNA 3＇end is one of the three main steps of eukaryotic pre-mRNA processing. The prediction of polyadenylation sites in human DNA and mRNA sequences is very important for realizing the pre-mRNA processing and prediction of gene structure. This paper presents a machine learning method to predict polyadenylation signals （PASes） in human DNA and mRNA sequences. This method consists of three steps of feature manipulation： Generation, selection and integration of features. In the first step, new features are generated using k-gram nucleotide acid patterns. In the second step, a number of important features are selected by an entropy-based algorithm. In the third step, support vector machines are employed to recognize true PASes from a large number of candidates. At last, a mathematic model forms. When the sensitivity is 60%, the corresponding specificity is 71.67% on intron level, and 80.77% on exon level.