基于DNA序列信息的核小体位点预测

Prediction of nucleosome positioning in genomes based on DNA sequence information

核小体位点参与很多细胞活动并且在调控细胞进程中起重要作用,核小体位点的精确定位对生物医学和药物学有重大意义。该文结合机器学习算法和DNA序列信息建立一个新的核小体位点预测模型,目的是为了更快更精确地预测核小体位点。实验中使用K间隔核酸对组成特征和位置特异性核苷酸偏好特征将DNA序列进行编码,然后使用支持向量机对模型进行训练和分类预测。最后将该方法和目前主流的核小体位点预测模型在相同的数据集上做Jackknife测试,实验结果证明了该方法提高了核小体位点预测的正确率。

Nucleosome positioning participates in many cellular activities and plays essential roles in regulating cellular processes. The prediction of nucleosome positioning in DNA is significantly important for both basic biomedicine research and practical drug development. This paper proposed a computational-based method for fast and accurately predicting nucleosome positioning solely from primary DNA sequences. It introduced the composition k-spaced nucleic acid pairs feature and position-specific nucleotide propensity feature for encoding DNA sequences. Then it trained a support vector machine（ SVM） prediction model with the feature on the training dataset. It we compared the proposed predictor with existing nucleosome positioning predictors by performing rigorous Jackknife tests on benchmark datasets,and the experimental results demonstrated that the proposed predictor outperformed existing nucleosome positioning predictors and remarkably improved the prediction performances.