Ensemble learning prediction of protein–protein interactions using proteins functional annotations

Abstract

Protein-protein interactions are important for the majority of biological processes. A significant number of computational methods have been developed to predict protein-protein interactions using protein sequence, structural and genomic data. Vast experimental data is publicly available on the Internet, but it is scattered across numerous databases. This fact motivated us to create and evaluate new high-throughput datasets of interacting proteins. We extracted interaction data from DIP, MINT, BioGRID and IntAct databases. Then we constructed descriptive features for machine learning purposes based on data from Gene Ontology and DOMINE. Thereafter, four well-established machine learning methods: Support Vector Machine, Random Forest, Decision Tree and Naïve Bayes, were used on these datasets to build an Ensemble Learning method based on majority voting. In cross-validation experiment, sensitivity exceeded 80% and classification/prediction accuracy reached 90% for the Ensemble Learning method. We extended the experiment to a bigger and more realistic dataset maintaining sensitivity over 70%. These results confirmed that our datasets are suitable for performing PPI prediction and Ensemble Learning method is well suited for this task. Both the processed PPI datasets and the software are available at .

A Novel Feature Extraction Scheme with Ensemble Coding for Protein–Protein Interaction Prediction

“…We used the Yeast and Human protein interaction datasets derived by Indrajit Saha et al [27]. The dataset was composed of 3 datasets extracted from the DIP, MINT, BioGrid, and IntAct databases.…”

“…These computational methods can be roughly divided into sequence based [11,12,13,14,15,16,17,18,19], structure based [20,21,22,23,24], and function annotation based [25,26,27,28,29] methods with different coding methods, such as autocovariance (AC) [12], local descriptors (LD) [19], conjoint triad (CT) [11], Geary autocorrelation (GAC) [30], Moran autocorrelation (MAC) [31], and normalized Moreau–Broto autocorrelation (NMBAC) [32]. Sequence-based methods have the advantage of not requiring expensive and time-consuming processes to determine protein structures.…”

Peer Review #1 of "Multi-level machine learning prediction of protein–protein interactions in Saccharomyces cerevisiae (v0.1)"

“…The construction of high quality negative examples is very difficult. Common methods for generating negatives include drawing random pairs of biomolecules from all known proteins found in a specific organism (Saha et al, 2014), or only from the selected subset of the whole proteome, namely from the proteins occurring in positive examples (Chang et al, 2010). We strongly believe that such methods have their inherent drawbacks, because they ignore network properties of the underlying protein interactome.…”

Quo vadis1 computational analysis of PPI data or why the future isn't here yet

“…State of the art computational methods for the prediction of PPI combine information from different sources and have presented adequate classification performance. Recent approaches (Zhang et al, 2012 ; Saha et al, 2014 ; Theofilatos et al, 2014 ) have attempted to overcome the bottlenecks in this PPIs prediction, namely the definition of negative datasets, the feature selection, the class imbalance, the tradeoff between classification performance and interpretability, missing features values and the calculation of a confidence score for every PPI. The advancements on the computational prediction and scoring of PPI enabled the construction of binary PPI networks with increased coverage on the full interactome.…”