Predicting Binding within Disordered Protein Regions to Structurally Characterised Peptide-Binding Domains

Abstract

Disordered regions of proteins often bind to structured domains, mediating interactions within and between proteins. However, it is difficult to identify a priori the short disordered regions involved in binding. We set out to determine if docking such peptide regions to peptide binding domains would assist in these predictions.We assembled a redundancy reduced dataset of SLiM (Short Linear Motif) containing proteins from the ELM database. We selected 84 sequences which had an associated PDB structures showing the SLiM bound to a protein receptor, where the SLiM was found within a 50 residue region of the protein sequence which was predicted to be disordered. First, we investigated the Vina docking scores of overlapping tripeptides from the 50 residue SLiM containing disordered regions of the protein sequence to the corresponding PDB domain. We found only weak discrimination of docking scores between peptides involved in binding and adjacent non-binding peptides in this context (AUC 0.58).Next, we trained a bidirectional recurrent neural network (BRNN) using as input the protein sequence, predicted secondary structure, Vina docking score and predicted disorder score. The results were very promising (AUC 0.72) showing that multiple sources of information can be combined to produce results which are clearly superior to any single source.We conclude that the Vina docking score alone has only modest power to define the location of a peptide within a larger protein region known to contain it. However, combining this information with other knowledge (using machine learning methods) clearly improves the identification of peptide binding regions within a protein sequence. This approach combining docking with machine learning is primarily a predictor of binding to peptide-binding sites, and is not intended as a predictor of specificity of binding to particular receptors.

DISOPRED3: precise disordered region predictions with annotated protein-binding activity

“…, 2012 ) and MFSPSSMpred ( Fang et al. , 2013 ), which are available for download or online use; PepBindPred ( Khan et al. , 2013 ) was not tested, given the running times needed for molecular dynamics simulations.…”

“…, 2013 ) considers only sequence profiles, which are pre-processed to enhance the signal of local conservation within the fast evolving landscape of IDR sequences. Finally, PepBindPred ( Khan et al. , 2013 ) first attempts to estimate the binding affinity of tripeptides from the input sequence to a library of known protein-binding domains, and then feeds these data to a bidirectional recurrent neural network along with additional features.…”

High-throughput prediction of RNA, DNA and protein binding regions mediated by intrinsic disorder

“…They include BindN+ ( 62 ) and RNABindR 2.0 ( 38 ) for the RNA-binding, and BindN+ and DNABR ( 63 ) for the DNA-binding. We also compared with the three predictors of the disordered protein–protein interacting residues: MoRFpred ( 26 ), DISOPRED3 ( 27 ) and ANCHOR ( 24 ); we did not include PepBindPred ( 25 ) due to the relatively long runtime required for the molecular dynamics simulations used by this method.…”

“…However, these predictions were performed at the whole protein level, were based on predicted disordered regions and assumed that the predicted IDRs contribute toward the GO annotations of the corresponding protein. The ANCHOR ( 24 ) and PepBindPred ( 25 ) methods that predict protein–protein binding residues located in IDRs and MoRFpred ( 26 ) and DISOPRED3 ( 27 ) methods that find short protein-binding regions (up to 25 consecutive residues) in IDRs that are involved in molecular recognition were also developed. These attempts suggest that functions of IDRs are predictable from the protein sequence.…”

Computational Prediction of MoRFs, Short Disorder-to-order Transitioning Protein Binding Regions

“…A collection of over 3000 SLiMs curated from literature is available in the ELM resource [ 43 , 45 ]. The two SLiM predictors, PepBindPred [ 105 ] and SLiMPred [ 106 ], utilize machine learning-derived neural network models. PepBindPred's model was derived using training datasets of SLiMs that were filtered to be embedded within putative IDRs, representing a motif-associated subpopulation of MoRFs.…”