为实现在海量的被测序DNA序列中快速、准确的定位核小体, 解决传统人工实验法和被提出的一些计算方法耗时长和准确率低等问题, 迫切需要设计一种快速有效的核小体自动化定位方法.在基于伪核苷酸K-联体特征提取的基础上构造样本集的特征向量, 提出在TensorFlow框架下利用卷积神经网络 (CNN) 构建核小体定位的网络预测模型.在预测模型上分别对智人、线虫和果蝇3个基准数据集进行交叉验证测试, 预测准确率分别为88.21％、89.19％、85.07％, 实验结果表明, 该预测模型性能高于目前已有预测模型.%To locate nucleosomes quickly and accurately in a large number of sequenced DNA sequences, and to solve the problems of long time-consuming and low accuracy of traditional manual experiments and some computational methods proposed, it is urgent to design a fast and effective automatic nucleosome localization method.Therefore, the feature vectors of the sample set were constructed based on the feature extraction of pseudo K-tuple nucleotide composition, and the network prediction model of nucleosome localization was constructed using convolutional neural network (CNN) in the framework of TensorFlow.Cross-validation tests were carried out on three benchmark data sets of Homo sapiens, Nematodes and Drosophila respectively.The prediction accuracy is 88.21％, 89.19％ and 85.07％respectively.Experimental results show that the performance of the prediction model constructed is obviously better than the existing prediction model.
展开▼
