|
Abstract:
|
In the current climate of high-throughput computational biology, the inference of a protein's function from related covariates, such as protein-protein interaction (PPI) relations, has become a canonical problem. Most existing technologies pursue this task as a Gene Ontology (GO) term-based classification problem. However, ontology structures are essentially hierarchies, with certain top to bottom annotation rules. We propose a probabilistic framework to integrate relational data, in the form of a PPI network, and the GO hierarchy, and offer two classifiers classifying GO terms marginally and jointly. Efficient dynamic programming algorithms for their computation are derived. We apply and evaluate our model in the yeast Saccharomyces cerevisiae for a whole-genome protein function prediction. It is found that substantial improvements may be obtained over non-hierarchical methods.
|
