RNA二级结构预测是计算分子生物学中的一个重要领域.本文介绍了RNA二级结构的预测方法,包括该问题的数学模型、主要算法思想以及每种算法对应的软件.在tRNA和RNase P RNA数据库中随机选取了几组样例对目前主要的7种软件进行测试,同时对...RNA二级结构预测是计算分子生物学中的一个重要领域.本文介绍了RNA二级结构的预测方法,包括该问题的数学模型、主要算法思想以及每种算法对应的软件.在tRNA和RNase P RNA数据库中随机选取了几组样例对目前主要的7种软件进行测试,同时对每种软件的优缺点进行了详细比较.实验证明,当存在同源序列时,Pfold的效果优于其它软件.最后,在总结分析现有算法的基础上探讨了该领域进一步的研究方向.展开更多
Pms1, a locus for photoperiod sensitive genic male sterility in rice, was identified and mapped to chromosome 7 in previous studies. Here we report an effort to identify the candidate genes for Pms1 by comparative seq...Pms1, a locus for photoperiod sensitive genic male sterility in rice, was identified and mapped to chromosome 7 in previous studies. Here we report an effort to identify the candidate genes for Pms1 by comparative sequencing of BAC clones from two cultivars Minghui 63 and Nongken 58, the parents for the initial mapping population. Annotation and comparison of the sequences of the two clones resulted in a total of five potential candidates which should be functionally tested. We also conducted com-parative analysis of sequences of these two cultivars with two other cultivars, Nipponbare and 93-11, for which sequence data were available in public databases. The analysis revealed large differences in sequence composition among the four genotypes in the Pms1 region primarily due to retroelement activity leading to rapid recent growth and divergence of the genomes. High levels of polymorphism in the forms of indels and SNPs were found both in intra- and inter-subspecific comparisons. Dating analysis using LTRs of the retroelements in this region showed that the substitution rate of LTRs was much higher than reported in the literature. The results provided strong evidence for rapid genomic evolution of this region as a consequence of natural and artificial selection.展开更多
为了解决传统机器学习对电能质量分类时人工选择特征困难导致分类效果不好的问题,本文提出了一种高度比较时间序列分析(highly comparative time series analysis,HCTSA)结合BP神经网络的分类方法,该方法既可以提取大量特征又可以筛选特...为了解决传统机器学习对电能质量分类时人工选择特征困难导致分类效果不好的问题,本文提出了一种高度比较时间序列分析(highly comparative time series analysis,HCTSA)结合BP神经网络的分类方法,该方法既可以提取大量特征又可以筛选特征,不需要人工选择特征指标。首先,使用大量算法数据库从时间序列中提取数千个可解释特征;然后,再利用线性分类器进行前向特征选择,选择出类结构中信息量最大的特征,即为最优特征。最后,用最优特征重新对扰动样本提取特征,并用BP神经网络分类。仿真结果表明,HCTSA-BP法的分类准确率达到了97.3%,比传统的小波-BP法高了8.7%。展开更多
文摘RNA二级结构预测是计算分子生物学中的一个重要领域.本文介绍了RNA二级结构的预测方法,包括该问题的数学模型、主要算法思想以及每种算法对应的软件.在tRNA和RNase P RNA数据库中随机选取了几组样例对目前主要的7种软件进行测试,同时对每种软件的优缺点进行了详细比较.实验证明,当存在同源序列时,Pfold的效果优于其它软件.最后,在总结分析现有算法的基础上探讨了该领域进一步的研究方向.
基金Supported by the National Program of High Technology Development of China (Grant No. 2006AA10A103)the National Natural Science Foundation of China (Grant No. 30321005)
文摘Pms1, a locus for photoperiod sensitive genic male sterility in rice, was identified and mapped to chromosome 7 in previous studies. Here we report an effort to identify the candidate genes for Pms1 by comparative sequencing of BAC clones from two cultivars Minghui 63 and Nongken 58, the parents for the initial mapping population. Annotation and comparison of the sequences of the two clones resulted in a total of five potential candidates which should be functionally tested. We also conducted com-parative analysis of sequences of these two cultivars with two other cultivars, Nipponbare and 93-11, for which sequence data were available in public databases. The analysis revealed large differences in sequence composition among the four genotypes in the Pms1 region primarily due to retroelement activity leading to rapid recent growth and divergence of the genomes. High levels of polymorphism in the forms of indels and SNPs were found both in intra- and inter-subspecific comparisons. Dating analysis using LTRs of the retroelements in this region showed that the substitution rate of LTRs was much higher than reported in the literature. The results provided strong evidence for rapid genomic evolution of this region as a consequence of natural and artificial selection.
文摘为了解决传统机器学习对电能质量分类时人工选择特征困难导致分类效果不好的问题,本文提出了一种高度比较时间序列分析(highly comparative time series analysis,HCTSA)结合BP神经网络的分类方法,该方法既可以提取大量特征又可以筛选特征,不需要人工选择特征指标。首先,使用大量算法数据库从时间序列中提取数千个可解释特征;然后,再利用线性分类器进行前向特征选择,选择出类结构中信息量最大的特征,即为最优特征。最后,用最优特征重新对扰动样本提取特征,并用BP神经网络分类。仿真结果表明,HCTSA-BP法的分类准确率达到了97.3%,比传统的小波-BP法高了8.7%。
