Using hidden Markov models to predict DNA-binding proteins with sequence and structure information

研究成果: Article同行評審

6 引文 斯高帕斯(Scopus)


In the post-genome period, the protein domain structures are published rapidly, but they have not been studied comprehensively. To figure out the cell function, the protein–DNA interactions decrypt the protein domain structures in recent research. Several machine-learning based methods are applied to the issue; however, they are not efficient to translate the tertiary structure characteristics of proteins into appropriate features for predicting the DNA-binding proteins. In this work, a novel machine-learning approach based on hidden Markov models identifies the characteristics of DNA-binding proteins with their amino acid sequences and tertiary structures. After we distill the features from DNA-binding proteins, a support vector machine based classifier predicts general DNA-binding proteins with the accuracy of 88.45 % through fivefolds cross-validation. Furthermore, we construct a response element specific classifier for predicting response element specific DNA-binding proteins, and the performance achieves the precision of 96.57 % with recall rate as 88.83 % in average. To verify the prediction of DNA-binding proteins, we used the DNA-binding proteins from MCF-7 that are likely to bind with estrogen response elements (ERE), and the results show that our methods can apply to practice.

頁(從 - 到)2365-2376
期刊Soft Computing
出版狀態Published - 2014 1月 1

All Science Journal Classification (ASJC) codes

  • 理論電腦科學
  • 軟體
  • 幾何和拓撲


深入研究「Using hidden Markov models to predict DNA-binding proteins with sequence and structure information」主題。共同形成了獨特的指紋。