Semi-supervised protein function prediction via sequential linear neighborhood propagation
ICIC'11 Proceedings of the 7th international conference on Intelligent Computing: bio-inspired computing and applications
Predicting Protein Function by Multi-Label Correlated Semi-Supervised Learning
IEEE/ACM Transactions on Computational Biology and Bioinformatics (TCBB)
Hi-index | 0.00 |
Protein function prediction represents a fundamental challenge in bioinformatics. The increasing availability of proteomics network data has enabled the development of several approaches that exploit the information encoded in networks in order to infer protein function. In this paper we introduce a new algorithm based on the concept of topological overlap between nodes of the graph, which addresses the problem of the classification of partially labeled protein interaction networks. The proposed approach is tested on the yeast interaction map and compared with two current state-of-the-art algorithms. Cross-validation experiments provide evidence that the proposed method represents a competitive alternative in a wide range of experimental conditions and also that, in many cases, it provides enhanced predictive accuracy.