一种预测判断蛋白质与DNA相互作用位点的新方法

A new method to predict Protein-DNA binding sites

摘要:

蛋白质-DNA相互作用位点在各类生理生化反应中扮演重要角色.本论文旨在构建一种可以准确预测“相互作用位点”的方法：PdDNA，其内容主要包括支持向量机和序列匹配器.支持向量机通过提取相互作用位点中心残基的特征进行训练并分类，序列匹配器则通过蛋白质特征矩阵（PSSM）对氨基酸序列进行相关性评估，对二者结果进行归一化整合，得到最终的预测结果.利用公开数据集PDNA_62，我们的PdDNA预测准确率为86.87%.为进一步验证PdDNA可靠性，我们还自建了PDNA_224数据集，其预测准确率为83.07%，处于较高水平.因此PdDNA是一种有效的“蛋白质-DNA相互作用位点”预测方法.

Abstract:

Protein-DNA binding sites play an important role in various physiological and biochemical reactions. In this paper, we establish a special method and algorithm based on Bioinformatics to forecast Protein-DNA binding sites, we call it PdDNA. According to our method we have 2 mainly algorithm: SVM-based predictor and sequence-based predictor. SVM-based predictor is trained and classified by extracting features of central residues at binding sites, and sequence-based predictor scores amino acid sequences for correlation by Position-Specific Scoring Matrix(PSSM). Normalization and integration of the two results to obtain the final forecast. According to our algorithm, it predicts DNA-binding sites with 86.87% accuracy when tested on PDNA_62 dataset. Otherwise, we established PDNA_224 data set, and PdDNA also has 83.07% accuracy at a high level. Therefore, PdDNA is an effective method for predicting "Protein-DNA binding sites ".