论文部分内容阅读
针对利用经典的随机上下文无关文法(SCFG)等模型对RNA(R ibonucle ic ac id)二级结构进行预测时,存在计算复杂性问题,该文给出了RNA二级结构的“新二级结构单元标签”(N SSEL)表示,相应提出了一种新的RNA二级结构预测的神经网络方法。这种二级结构的N SSEL表示格式很容易转换成常用的CT格式。基于tRNA数据集的实验表明,在完全相同的训练与测试数据集下,该方法,较之性能最好的B JK与BK 2等SCFG模型,其预测精度与相关系数都有所提高,证明了所提方法的可行性与有效性。由于神经网络启发式方法不存在计算时间复杂性问题,因此可望将此法用于预测SCFG等算法难以处理的大于1 000个碱基的长RNA序列的折叠问题。
In order to predict the RNA secondary structure using the classical random-independent context-free grammar (SCFG) model, the computational complexity exists. This paper presents a new secondary structure of RNA secondary structure N SSEL "indicates that a new neural network prediction method of RNA secondary structure is proposed accordingly. This secondary structure N SSEL representation format is easily converted to the common CT format. Experiments based on tRNA datasets show that this method, compared with the best performing SCFG model, such as BJK and BK 2, has an improved prediction accuracy and correlation coefficient under the same training and testing dataset. It is proved that The feasibility and effectiveness of the proposed method. Because there is no computational time complexity in neural network heuristic, it is expected that this method can be used to predict the folding problem of long RNA sequences longer than 1 000 bases which is difficult to be solved by SCFG and other algorithms.