Protein-RNA interface residue prediction using machine learning: an assessment of the state of the art

Abstract

BackgroundRNA molecules play diverse functional and structural roles in cells. They function as messengers for transferring genetic information from DNA to proteins, as the primary genetic material in many viruses, as catalysts (ribozymes) important for protein synthesis and RNA processing, and as essential and ubiquitous regulators of gene expression in living organisms. Many of these functions depend on precisely orchestrated interactions between RNA molecules and specific proteins in cells. Understanding the molecular mechanisms by which proteins recognize and bind RNA is essential for comprehending the functional implications of these interactions, but the recognition ‘code’ that mediates interactions between proteins and RNA is not yet understood. Success in deciphering this code would dramatically impact the development of new therapeutic strategies for intervening in devastating diseases such as AIDS and cancer. Because of the high cost of experimental determination of protein-RNA interfaces, there is an increasing reliance on statistical machine learning methods for training predictors of RNA-binding residues in proteins. However, because of differences in the choice of datasets, performance measures, and data representations used, it has been difficult to obtain an accurate assessment of the current state of the art in protein-RNA interface prediction.ResultsWe provide a review of published approaches for predicting RNA-binding residues in proteins and a systematic comparison and critical assessment of protein-RNA interface residue predictors trained using these approaches on three carefully curated non-redundant datasets. We directly compare two widely used machine learning algorithms (Naïve Bayes (NB) and Support Vector Machine (SVM)) using three different data representations in which features are encoded using either sequence- or structure-based windows. Our results show that (i) Sequence-based classifiers that use a position-specific scoring matrix (PSSM)-based representation (PSSMSeq) outperform those that use an amino acid identity based representation (IDSeq) or a smoothed PSSM (SmoPSSMSeq); (ii) Structure-based classifiers that use smoothed PSSM representation (SmoPSSMStr) outperform those that use PSSM (PSSMStr) as well as sequence identity based representation (IDStr). PSSMSeq classifiers, when tested on an independent test set of 44 proteins, achieve performance that is comparable to that of three state-of-the-art structure-based predictors (including those that exploit geometric features) in terms of Matthews Correlation Coefficient (MCC), although the structure-based methods achieve substantially higher Specificity (albeit at the expense of Sensitivity) compared to sequence-based methods. We also find that the expected performance of the classifiers on a residue level can be markedly different from that on a protein level. Our experiments show that the classifiers trained on three different non-redundant protein-RNA interface datasets achieve comparable cross-validation performance. However...

Insights into the Mechanism of Ribosomal Incorporation of Mammalian L13a Protein during Ribosome Biogenesis

“…1a, left and middle) and probably serves to anchor the protein to the rRNA. Interestingly, a Web-based server, RNABindR (http://einstein.cs.iastate.edu /RNABindR/) (20), which employs a distance cutoff to predict amino acids as probable candidates to contact RNA using solved ribonucleoprotein (RNP) structures available in the Protein Data Bank (PDB), also identified Arg68 of L13a as a potential RNA binding candidate within a region comprised of residues 53 to 75. The RNABindR algorithm also predicted a second motif within the region spanning Arg169 to Gln173 as a potential RNA binding site.…”

Computational Tools for Investigating RNA-Protein Interaction Partners

“…As a step toward deciphering the rules that govern recognition specificity in RNAprotein interfaces, many computational methods (both sequencebased and structure-based) have been developed for predicting RNAbinding residues in proteins. Three recent reviews have summarized and compared these methods [21,27,28], which we will not reconsider here. With one exception, all published methods for predicting RNAbinding residues in a protein of interest do not take into account the specific RNA partner with which it interacts (i.e.…”

Models of RNA Interaction from Experimental Datasets: Framework of Resilience

“…We have determined the secondary structure of all RefSeq transcripts to predict single-stranded regions using RNAfold [24], while others have used RNA structure predictors (RNAplfold in Refs. [25,26]) in a pooling predictor using machine learning [27]. Additionally, nucleotide solvent-accessibility in RNA structures could be estimated by the neural network method of Singh [22] using models of window size 3 nt, which could be expanded to 5-9 nt windows for k length.…”