Predicting RNA-binding sites of proteins using support vector machines and evolutionary information

Abstract

Background: RNA-protein interaction plays an essential role in several biological processes, such as protein synthesis, gene expression, posttranscriptional regulation and viral infectivity. Identification of RNA-binding sites in proteins provides valuable insights for biologists. However, experimental determination of RNA-protein interaction remains time-consuming and laborintensive. Thus, computational approaches for prediction of RNA-binding sites in proteins have become highly desirable. Extensive studies of RNA-binding site prediction have led to the development of several methods. However, they could yield low sensitivities in trade-off for high specificities.

Annotating the protein-RNA interaction sites in proteins using evolutionary information and protein backbone structure

“…For the study of RNA-binding sites, the three common data bases: RBP86 dataset, RBP107 dataset and RBP109 dataset which have been widely used by many researchers (Ahmad et al, 2003;Cheng et al, 2008;Jeong et al, 2004;Kumar et al, 2008;Terribilini et al, 2006;Wang et al, 2011aWang et al, , 2008Wang and Brown, 2006) are used to evaluate the performance of our method.…”

“…Evolutionary information in the form of position-specific scoring matrix (PSSM) profile had been used to identify the function of proteins (Cheng et al, 2008(Cheng et al, , 2007aDu and Li, 2008;Hayat and Khan, 2012;Kumar et al, 2008;Kuznetsov et al, 2006;Li et al, 2010aLi et al, , 2010bPu et al, 2007;Chou, 2007, 2009;Shu et al, 2008;Wang et al, 2011aWang et al, , 2008Xiong et al, 2010). In this work, the evolutionary information is expressed by the PSSM profiles.…”

“…PRNA (Liu et al, 2010) generated a random forest (RF) method to predict RNA-binding amino acid residues in proteins. Recently, RNA-binding sites in proteins were predicted by combining the improved Position-specific scoring matrix (PSSM) with a SVM classifier (Cheng et al, 2008;Wang et al, 2011a). However, the protein backbone structure between protein residue and RNA nucleotide have not been considered in most of the methods.…”

Improved feature-based prediction of SNPs in human cytochrome P450 enzymes

“…When the size of sliding window w equals 11, V i includes 5 rows upstream and 5 rows downstream to the ith row (in the bigger red rectangle of the original PSSM profile). For the N-terminal and C-terminal of a protein, ZERO vectors, which consist of 20 zero elements, are appended to the head or tail of a PSSM profile (Cheng et al, 2008). Finally, the smoothed PSSM was normalized by the sigmoid function f (x) = 1/(1 + e −x ).…”

Predicting protein-binding RNA nucleotides using the feature-based removal of data redundancy and the interaction propensity of nucleotide triplets

“…Preparing enough training data is crucial for the success of a supervised learning method but the training data should be non-redundant and representative because redundant data can cause a prediction bias and make a prediction model take more time and space to train due to the increased size of a training dataset. Many learning approaches to predict RNA-binding residues construct a training dataset based on the similarity of protein sequences [3][4][5][13][14][15]. When similar sequences are eliminated from a dataset, their binding information is also lost, which would otherwise be valuable for predicting binding sites.…”