Transient protein-protein interface prediction: datasets, features, algorithms, and the RAD-T predictor

Abstract

BackgroundTransient protein-protein interactions (PPIs), which underly most biological processes, are a prime target for therapeutic development. Immense progress has been made towards computational prediction of PPIs using methods such as protein docking and sequence analysis. However, docking generally requires high resolution structures of both of the binding partners and sequence analysis requires that a significant number of recurrent patterns exist for the identification of a potential binding site. Researchers have turned to machine learning to overcome some of the other methods’ restrictions by generalising interface sites with sets of descriptive features. Best practices for dataset generation, features, and learning algorithms have not yet been identified or agreed upon, and an analysis of the overall efficacy of machine learning based PPI predictors is due, in order to highlight potential areas for improvement.ResultsThe presence of unknown interaction sites as a result of limited knowledge about protein interactions in the testing set dramatically reduces prediction accuracy. Greater accuracy in labelling the data by enforcing higher interface site rates per domain resulted in an average 44% improvement across multiple machine learning algorithms. A set of 10 biologically unrelated proteins that were consistently predicted on with high accuracy emerged through our analysis. We identify seven features with the most predictive power over multiple datasets and machine learning algorithms. Through our analysis, we created a new predictor, RAD-T, that outperforms existing non-structurally specializing machine learning protein interface predictors, with an average 59% increase in MCC score on a dataset with a high number of interactions.ConclusionCurrent methods of evaluating machine-learning based PPI predictors tend to undervalue their performance, which may be artificially decreased by the presence of un-identified interaction sites. Changes to predictors’ training sets will be integral to the future progress of interface prediction by machine learning methods. We reveal the need for a larger test set of well studied proteins or domain-specific scoring algorithms to compensate for poor interaction site identification on proteins in general.

Analysis of a Therapeutic Antibody Cocktail Reveals Determinants for Cooperative and Broad Ebolavirus Neutralization

“…All model figures were generated using UCSF Chimera (Pettersen et al, 2004). To estimate change in GP solvent exclusion upon Fabs binding, GP1,2 from an unliganded GP structure (PDB: 5JQ3), and GP1,2 bound to the respective Fabs were used to create a surface map in UCSF chimera (Pettersen et al, 2004) and then normalized to Gly-X-Gly tripeptides (Bendell et al, 2014) to determine a normalized solvent excluded surface (SES).…”

pH Dependence of Charge Multipole Moments in Proteins

“…We obtain the RSA of each AA in a protein with the help of the UCSF Chimera software [53,54] by normalizing the solvent-exposed surface area of each residue in the protein structure with the surface area of the same type of residue in a reference state [55]. The classification of AA residues as "buried" or "exposed" is then done on the basis of an RSA cut-off c, which typically ranges between 5% and 30% [56][57][58][59].…”

Utilizing knowledge base of amino acids structural neighborhoods to predict protein-protein interaction sites

“…The MCC values of the other methods are taken from [5]. The comparison with PredUs [4], PrISE [23], RAD-T [6] and MetaPPISP [22] are summarized in Table 4. Performance of those methods, which also includes precision, recall, accuracy and F1 measure, are borrowed from [3,6,23].…”

“…The trained model is subsequently used when an unknown protein needs to be characterized. A number of descriptors have been utilized for the purpose of PPI identification, such as hydrophobicity [5], energy of solvatation [6], propensity [5] or RASA (Relative Solvent Accessible Surface Area) [3][4][5][6], with RASA being especially popular [7]. As for machine learning approaches, the best performing methods utilize Support Vector Machines (SVM) [3,5], Neural networks [8], Decision trees [6] or Conditional Random Fields (CRF) [9,10].…”