A novel one-class SVM based negative data sampling method for reconstructing proteome-wide HTLV-human protein interaction networks

Abstract

Protein-protein interaction (PPI) prediction is generally treated as a problem of binary classification wherein negative data sampling is still an open problem to be addressed. The commonly used random sampling is prone to yield less representative negative data with considerable false negatives. Meanwhile rational constraints are seldom exerted on model selection to reduce the risk of false positive predictions for most of the existing computational methods. In this work, we propose a novel negative data sampling method based on one-class SVM (support vector machine, SVM) to predict proteome-wide protein interactions between HTLV retrovirus and Homo sapiens, wherein one-class SVM is used to choose reliable and representative negative data, and two-class SVM is used to yield proteome-wide outcomes as predictive feedback for rational model selection. Computational results suggest that one-class SVM is more suited to be used as negative data sampling method than two-class PPI predictor, and the predictive feedback constrained model selection helps to yield a rational predictive model that reduces the risk of false positive predictions. Some predictions have been validated by the recent literature. Lastly, gene ontology based clustering of the predicted PPI networks is conducted to provide valuable cues for the pathogenesis of HTLV retrovirus.

Multi-label ℓ2-regularized logistic regression for predicting activation/inhibition relationships in human protein-protein interaction networks

“…In the scenario of traditional supervised learning, both the target instance and the homolog instance are predicted to one class label. The predicted class labels of the two instances can be easily combined into a final label by comparing their decision values2328. For instance, the target instance is predicted to the label , while the homolog instance is predicted to the label .…”

“…The annotations of these three aspects of genes or gene products are provided in terms of GO terms in the GOA database18. Recently GO terms have been successfully used as features to predict protein-protein interactions1920212223. There are two effective approaches to exploit GO terms for representing protein pairs.…”

“…There are two effective approaches to exploit GO terms for representing protein pairs. One approach is to exploit the shared GO terms between two proteins and construct explicit binary feature vectors as the inputs of machine learning methods20212223, and the other approach is to measure the similarity between GO terms in GO DAG (directed acyclic graph) and construct an implicit kernel matrix as the input of kernel methods19.…”

“…One method is to directly use the GO terms extracted from the GOA database alone212223, and the other method is to incorporate the ancestor GO terms of each GO term concerned20. Incorporation of ancestor GO terms surely adds useful information to the training data and thus improves the model performance.…”

Towards Data Analytics of Pathogen-Host Protein-Protein Interaction: A Survey

“…The other challenge in dealing with the negative PPIs data is its selection strategy. Both random sampling and a novel negative data sampling method based on one-class SVM have been compared in [56].…”