Accurate prediction of RNA-binding protein residues with two discriminative structural descriptors

Abstract

BackgroundRNA-binding proteins participate in many important biological processes concerning RNA-mediated gene regulation, and several computational methods have been recently developed to predict the protein-RNA interactions of RNA-binding proteins. Newly developed discriminative descriptors will help to improve the prediction accuracy of these prediction methods and provide further meaningful information for researchers.ResultsIn this work, we designed two structural features (residue electrostatic surface potential and triplet interface propensity) and according to the statistical and structural analysis of protein-RNA complexes, the two features were powerful for identifying RNA-binding protein residues. Using these two features and other excellent structure- and sequence-based features, a random forest classifier was constructed to predict RNA-binding residues. The area under the receiver operating characteristic curve (AUC) of five-fold cross-validation for our method on training set RBP195 was 0.900, and when applied to the test set RBP68, the prediction accuracy (ACC) was 0.868, and the F-score was 0.631.ConclusionsThe good prediction performance of our method revealed that the two newly designed descriptors could be discriminative for inferring protein residues interacting with RNAs. To facilitate the use of our method, a web-server called RNAProSite, which implements the proposed method, was constructed and is freely available at http://lilab.ecust.edu.cn/NABind.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-1110-x) contains supplementary material, which is available to authorized users.

Ensemble Architecture for Prediction of Enzyme‐ligand Binding Residues Using Evolutionary Information

“…Further, based on earlier studies, the impact of smoothing and condensing PSSM was explored. This was done using biochemical factors such as reported for identification of Flavin Adenine Dinucleotide (FAD), Nicotinamide Adenine Dinucleotide (NAD), Adenosine triphosphate (ATP) and Heme . For a residue R i : …”

“…As extensively reviewed in many studies, some of these are applicable for one or more groups of ligands,, and are therefore, categorized under the general approaches. Some others are more specific, with applicability in protein interactions with DNA, RNA, Heme, zinc, vitamin, mannose, etc. Developing a widely applicable approach requires careful considerations of various underlying assumptions with regards to the diverse functional architecture for their eventual use .…”

Protein–RNA interactions: structural biology and computational modeling techniques

“…The common sequence features used are position-specific scoring matrices (PSSMs) and amino acid propensities, which quantify the propensity for an amino acid to be in a binding site (Murakami et al 2010). Frequently used structural features include the solvent accessibility of amino acids, surface electrostatic potentials, and geometrical properties such as shape (Sun et al 2016). The prediction techniques that use these features include machine learning (ML), template methods, and scoring methods.…”

“…Random forests (e.g., Barik et al 2015), support vector machines (e.g., Maetschke and Yuan 2009) and Naïve Bayes classifiers (e.g., Terribilini et al 2007) have all been used to predict of RNA-binding sites. One of the most recent methods (RNAProSite) constructs a random forest classifier using electrostatic surface potential and a triplet interface propensity (Sun et al 2016). Other methods have used ensemble learning, in which multiple ML classifiers are independently trained, and then combined to make predictions for target proteins (e.g., Ren and Shen 2015).…”

To improve the predictions of binding residues with DNA, RNA, carbohydrate, and peptide via multiple-task deep neural networks

“…For DNA-binding residue predictions, representative methods include TargetDNA (Jun Hu, 2017) and HMMBinder (Rianon Zaman, 2017) based on SVM, CNNsites (Wang, 2016) based on CNN network, DRNApred (Jing Yan, 2017) for accurately predicting and discriminating between DNA-and RNA-binding residues, and SPOT-DNA-Seq (Zhao, et al, 2014) based on alignments with known DNA-binding proteins. For RNA-binding residue predictions, there are RNAProSite (Meijian Sun, 2016) based on the random forest classifier and PredRBR (Yongjun Tang, 2017) based on the gradient tree boosting. For predictions of carbohydrate binding residues, the common idea is to find residues frequently observed on the sugar interface(Ghazaleh Taherzadeh, 2016) (Sujatha M S, 2004).…”