Protein-protein interaction based on pairwise similarity

Zaki, Nazar; Lazarova-Molnar, Sanja; El-Hajj, Wassim; Campbell, Piers

doi:10.1186/1471-2105-10-150

Cited by 57 publications

(33 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…First, several effective computational methods identify protein binding sites, mainly based on three-dimensional structural information [15] and protein sequences information [18,13,14]. However, 14-3-3 phosphopeptide binders only have six meaningful positions in binding motif sequences, and the existing stateof-the-art binding sites prediction methods must be not suitable for this issue.…”

Section: Introductionmentioning

confidence: 99%

Learning from real imbalanced data of 14-3-3 proteins binding specificity

Tang

Guo

2016

Neurocomputing

View full text Add to dashboard Cite

a b s t r a c tThe 14-3-3 proteins are a highly conserved family of homodimeric and heterodimeric molecules, expressed in all eukaryotic cells. In human cells, this family consists of seven distinct but highly homologous 14-3-3 isoforms. 14-3-3s is the only isoform directly linked to cancer in epithelial cells, which is regulated by major tumor suppressor gene. For each 14-3-3 isoform, we have 1000 peptide motifs with experimental binding affinity values. In this paper, we present a novel method for identifying peptide motifs binding to 14-3-3s isoform. First, we select nine physicochemical properties of amino acids to describe each peptide motif. We also use auto-cross covariance to extract correlative properties of amino acids in any two positions. Then, a similarity-based undersampling approach and a SMOTE-like oversampling approach are used to deal with imbalanced distribution of the known peptide motifs. Finally, we consider locally weighted regression to predict affinity values of peptide motifs, which combines the simplicity of linear least squares regression with the flexibility of nonlinear regression. Our method tests on the 1000 peptide motifs binding to seven 14-3-3 isoforms. On the 14-3-3s isoform, our method has overall Pearson-product-moment correlation coefficient (PCC) and the root mean squared error (RMSE) values of 0.83 and 258.31 for N-terminal sublibrary, and 0.80 and 250.89 for C-terminal sublibrary, respectively. We identify phosphopeptides that preferentially bind to 14-3-3s over other isoforms. Several positions on peptide motifs have the same amino acid as experimental substrate specificity of phosphopeptides binding to 14-3-3s. Our method is a fast and reliable computational method that can be used in peptide-protein binding identification in proteomics research.

show abstract

Section: Introductionmentioning

confidence: 99%

Learning from real imbalanced data of 14-3-3 proteins binding specificity

Tang

Guo

2016

Neurocomputing

View full text Add to dashboard Cite

show abstract

“…Moreover, amino acid (aa) sequences are the most universal protein feature and thus appear to be ideal traits for use in building methods of predicting PPIs that are applicable to all proteins [24]. In fact, many interesting and useful bioinformatics methods using primary sequences have been developed, and many methods include machine learning approaches [12], [15], [16], [24]–[36]. However, some of the methods based on ML and primary sequences have weaknesses, such as the building of negative datasets, the small number of examples and the large number of attributes (vectors) that code protein pairs.…”

Section: Introductionmentioning

confidence: 99%

The Development of a Universal In Silico Predictor of Protein-Protein Interactions

et al. 2013

View full text Add to dashboard Cite

Protein-protein interactions (PPIs) are essential for understanding the function of biological systems and have been characterized using a vast array of experimental techniques. These techniques detect only a small proportion of all PPIs and are labor intensive and time consuming. Therefore, the development of computational methods capable of predicting PPIs accelerates the pace of discovery of new interactions. This paper reports a machine learning-based prediction model, the Universal In Silico Predictor of Protein-Protein Interactions (UNISPPI), which is a decision tree model that can reliably predict PPIs for all species (including proteins from parasite-host associations) using only 20 combinations of amino acids frequencies from interacting and non-interacting proteins as learning features. UNISPPI was able to correctly classify 79.4% and 72.6% of experimentally supported interactions and non-interacting protein pairs, respectively, from an independent test set. Moreover, UNISPPI suggests that the frequencies of the amino acids asparagine, cysteine and isoleucine are important features for distinguishing between interacting and non-interacting protein pairs. We envisage that UNISPPI can be a useful tool for prioritizing interactions for experimental validation.

show abstract

“…We used these types of features for one data set in the experimental evaluation. Furthermore, informative features for predicting PPIs can be constructed from sequence information only, for example [40]. Each protein is represented by a vector of pairwise similarities against large subsequences of amino acids created by a shifting window which passes over concatenated protein training sequences.…”

Section: A Related Workmentioning

confidence: 99%

A Bayesian Framework for Combining Protein and Network Topology Information for Predicting Protein-Protein Interactions

Birlutiu

d’Alché‐Buc

Heskes

2015

IEEE/ACM Trans. Comput. Biol. and Bioinf.

View full text Add to dashboard Cite

Abstract-Computational methods for predicting proteinprotein interactions are important tools that can complement high-throughput technologies and guide biologists in designing new laboratory experiments. The proteins and the interactions between them can be described by a network which is characterized by several topological properties. Information about proteins and interactions between them, in combination with knowledge about topological properties of the network, can be used for developing computational methods that can accurately predict unknown protein-protein interactions. This paper presents a supervised learning framework based on Bayesian inference for combining two types of information: i) network topology information, and ii) information related to proteins and the interactions between them. The motivation of our model is that by combining these two types of information one can achieve a better accuracy in predicting protein-protein interactions, than by using models constructed from these two types of information independently.

show abstract

Protein-protein interaction based on pairwise similarity

Cited by 57 publications

References 28 publications

Learning from real imbalanced data of 14-3-3 proteins binding specificity

Learning from real imbalanced data of 14-3-3 proteins binding specificity

The Development of a Universal In Silico Predictor of Protein-Protein Interactions

A Bayesian Framework for Combining Protein and Network Topology Information for Predicting Protein-Protein Interactions

Contact Info

Product

Resources

About