The Evolutionary Conservation-based Analysis And Prediction For DNA-binding Residues

DNA-binding proteins play an indispensable role in various biological activities.The analysis and identification of DNA-binding residues(DBRs)is beneficial to better understanding of the mechanism of chromatin remodeling,gene regulation and expression.The employment of computational approaches can tangibly expedite such process and whilst provide instructive clues for wet-lab experiments.In terms of the evolutionary conservation,there are many statistical differences between DBRs and non-DBRs.Such differences play an important role in classifying DBRs and non-DBRs with machine learning methods.In the filed of DBRs prediction,traditional computational approaches usually construct static datasets to generate their knowledge-based models.Many homologous DNA-binding proteins are thus excluded so as to guarantee the generalization capability of the models.They can,however,potentially provide useful information for the investigation of protein–DNA interactions.In light of this,this study proposes a novel evolutionary conservation-based method to fill the gap of traditional machine learning patterns as well as to make some effort for better identification of DBRs.First,this study constructs a large-scale extensible sample pool to include useful samples.Then,this study uses evolutionary conservation-based features in forms of relative position specific score matrices and covariant evolutionary conservation descriptors to encode the feature space.Finally,this study designs a dynamic query-driven learning scheme to make more use of proteins with known structure and function.In comparison with traditional approaches,the introduction of dynamic models can improve the prediction performance to a great extent.Experimental results on the independent datasets prove that the proposed method has promising generalization capability.It is capable of producing decent predictions,which outperform many state-of-the-art methods.Moreover,this study also implements the method as an online webserver for the convenience of academic use.It is available at: http://www.inforstation.com/webservers/DQPred-DBR/.