Hierarchical Multilabel Protein Function Prediction Using Local Neural Networks

Protein function predictions are usually treated as classification problems where each function is regarded as a class label. However, different from conventional classification problems, they have some specificities that make the classification task more complex. First, the problem classes (protein functions) are usually hierarchically structured, with superclasses and subclasses. Second, proteins can be simultaneously assigned to more than one class in each hierarchical level, i.e., a protein can be classified into two or more paths of the hierarchical structure. This classification task is named hierarchical multilabel classification, and several methods have been proposed to deal with it. These methods, however, either transform the original problem into a set of simpler problems, loosing important information in this process, or employ complex internal mechanisms. Additionally, current methods have problems dealing with a high number of classes and also when predicting classes located in the deeper hierarchical levels, because the datasets become very sparse as their hierarchies are traversed toward the leaves. This paper investigates the use of local artificial neural networks for hierarchical multilabel classification, particularly protein function prediction. The proposed method was compared with state-of-the-art methods using several protein function prediction datasets. The experimental results suggest that artificial neural networks constitute a promising alternative to deal with hierarchical multilabel classification problems.