Improved multi-level protein–protein interaction prediction with semantic-based regularization

Saccà, Claudio; Teso, Stefano; Diligenti, Michelangelo; Passerini, Andrea

doi:10.1186/1471-2105-15-103

Cited by 19 publications

(20 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By applying a linear combination of the energy score, interface propensity, and residue conservation score, Liang et al [ 24 ] achieved decent coverage and accuracy. Zhang et al focused their effort on the interface conservation across structure space [ 25 ], while Saccà et al introduced multilevel (protein, domain, and residue) binding recognition using the Semantic Based Regularization Machine Learning framework [ 26 ]. It was shown that the three-dimensional structural information, either based on data from PDB [ 27 ] or obtained from homology modeling, produces a robust and efficient prediction of protein-protein interactions when applied with information on structural neighbors of queried proteins and Bayesian classifiers [ 28 ].…”

Section: Introductionmentioning

confidence: 99%

Heterodimer Binding Scaffolds Recognition via the Analysis of Kinetically Hot Residues

Perišić

2018

Pharmaceuticals

View full text Add to dashboard Cite

Physical interactions between proteins are often difficult to decipher. The aim of this paper is to present an algorithm that is designed to recognize binding patches and supporting structural scaffolds of interacting heterodimer proteins using the Gaussian Network Model (GNM). The recognition is based on the (self) adjustable identification of kinetically hot residues and their connection to possible binding scaffolds. The kinetically hot residues are residues with the lowest entropy, i.e., the highest contribution to the weighted sum of the fastest modes per chain extracted via GNM. The algorithm adjusts the number of fast modes in the GNM’s weighted sum calculation using the ratio of predicted and expected numbers of target residues (contact and the neighboring first-layer residues). This approach produces very good results when applied to dimers with high protein sequence length ratios. The protocol’s ability to recognize near native decoys was compared to the ability of the residue-level statistical potential of Lu and Skolnick using the Sternberg and Vakser decoy dimers sets. The statistical potential produced better overall results, but in a number of cases its predicting ability was comparable, or even inferior, to the prediction ability of the adjustable GNM approach. The results presented in this paper suggest that in heterodimers at least one protein has interacting scaffold determined by the immovable, kinetically hot residues. In many cases, interacting proteins (especially if being of noticeably different sizes) either behave as a rigid lock and key or, presumably, exhibit the opposite dynamic behavior. While the binding surface of one protein is rigid and stable, its partner’s interacting scaffold is more flexible and adaptable.

show abstract

Section: Introductionmentioning

confidence: 99%

Heterodimer Binding Scaffolds Recognition via the Analysis of Kinetically Hot Residues

Perišić

2018

Pharmaceuticals

View full text Add to dashboard Cite

show abstract

“…They constructed a classifier allowing for the information flow between the above three levels to improve the final prediction. This approach was further developed by Saccà et al (2014) , who introduced a different model of knowledge integration, and demonstrated its superiority on the previously used benchmarking data. Reported AUC values reached 0.80 for residues, 0.96 for domains and 0.82 for proteins.…”

Section: Introductionmentioning

confidence: 99%

Multi-level machine learning prediction of protein–protein interactions inSaccharomyces cerevisiae

Zubek

Tatjewski

Boniecki

et al. 2015

PeerJ

View full text Add to dashboard Cite

Accurate identification of protein–protein interactions (PPI) is the key step in understanding proteins’ biological functions, which are typically context-dependent. Many existing PPI predictors rely on aggregated features from protein sequences, however only a few methods exploit local information about specific residue contacts. In this work we present a two-stage machine learning approach for prediction of protein–protein interactions. We start with the carefully filtered data on protein complexes available for Saccharomyces cerevisiae in the Protein Data Bank (PDB) database. First, we build linear descriptions of interacting and non-interacting sequence segment pairs based on their inter-residue distances. Secondly, we train machine learning classifiers to predict binary segment interactions for any two short sequence fragments. The final prediction of the protein–protein interaction is done using the 2D matrix representation of all-against-all possible interacting sequence segments of both analysed proteins. The level-I predictor achieves 0.88 AUC for micro-scale, i.e., residue-level prediction. The level-II predictor improves the results further by a more complex learning paradigm. We perform 30-fold macro-scale, i.e., protein-level cross-validation experiment. The level-II predictor using PSIPRED-predicted secondary structure reaches 0.70 precision, 0.68 recall, and 0.70 AUC, whereas other popular methods provide results below 0.6 threshold (recall, precision, AUC). Our results demonstrate that multi-scale sequence features aggregation procedure is able to improve the machine learning results by more than 10% as compared to other sequence representations. Prepared datasets and source code for our experimental pipeline are freely available for download from: (open source Python implementation, OS independent).

show abstract

“…Gene locations were obtained from SGD; γ was set to 1. (ii) Similarly, protein complexes offer (noisy and incomplete) evidence about protein–protein interactions [22, 38]. We incorporated this information through a diffusion kernel K complex ( p,p ′ ) over the catalogue of yeast protein complexes [39].…”

Section: Resultsmentioning

confidence: 99%

“…Finally, the interaction predicate works on pairs of proteins, and thus requires a kernel between protein pairs . Following Saccà et al [22], we computed the pairwise kernel K pairwise (( p,p ′ ),( q,q ′ )) from the aggregate kernel K ( p,p ′ ) as follows: The pairwise kernel was also normalized and preconditioned.…”

Section: Resultsmentioning

confidence: 99%

“…OCELOT follows the same principles, but relies on Semantic Based Regularization, a different, sound structured-output method. SBR has previously been applied to multi-level PPI prediction [22]. Contrary to structured-output SVMs, SBR can be easily adapted to different prediction tasks by changing the consistency rules, as described in Methods.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Combining learning and constraints for genome-wide protein annotation

et al. 2019

Self Cite

View full text Add to dashboard Cite

Background The advent of high-throughput experimental techniques paved the way to genome-wide computational analysis and predictive annotation studies. When considering the joint annotation of a large set of related entities, like all proteins of a certain genome, many candidate annotations could be inconsistent, or very unlikely, given the existing knowledge. A sound predictive framework capable of accounting for this type of constraints in making predictions could substantially contribute to the quality of machine-generated annotations at a genomic scale. Results We present Ocelot , a predictive pipeline which simultaneously addresses functional and interaction annotation of all proteins of a given genome. The system combines sequence-based predictors for functional and protein-protein interaction (PPI) prediction with a consistency layer enforcing (soft) constraints as fuzzy logic rules. The enforced rules represent the available prior knowledge about the classification task, including taxonomic constraints over each GO hierarchy (e.g. a protein labeled with a GO term should also be labeled with all ancestor terms) as well as rules combining interaction and function prediction. An extensive experimental evaluation on the Yeast genome shows that the integration of prior knowledge via rules substantially improves the quality of the predictions. The system largely outperforms GoFDR, the only high-ranking system at the last CAFA challenge with a readily available implementation, when GoFDR is given access to intra-genome information only (as Ocelot ), and has comparable or better results (depending on the hierarchy and performance measure) when GoFDR is allowed to use information from other genomes. Our system also compares favorably to recent methods based on deep learning. Electronic supplementary material The online version of this article (10.1186/s12859-019-2875-5) contains supplementary material, which is available to authorized users.

show abstract

Improved multi-level protein–protein interaction prediction with semantic-based regularization

Cited by 19 publications

References 45 publications

Heterodimer Binding Scaffolds Recognition via the Analysis of Kinetically Hot Residues

Heterodimer Binding Scaffolds Recognition via the Analysis of Kinetically Hot Residues

Multi-level machine learning prediction of protein–protein interactions inSaccharomyces cerevisiae

Combining learning and constraints for genome-wide protein annotation

Contact Info

Product

Resources

About