Induction of Enzyme Classes from Biological Databases

Abstract

Bioinformatics is characterised by a growing diversity of large-scale databases containing information on genetics, proteins, metabolism and disease. It is widely agreed that there is an increasingly urgent need for technologies which can integrate these disparate knowledge sources. In this paper we propose that not only is machine learning a good candidate technology for such data integration, but Inductive Logic Programming, in particular, has strengths for handling the relational aspects of this task. Relations can be used to capture, in a single representation, not only biochemical reaction information but also protein and ligand structure as well as metabolic network information. Resources such as the Gene Ontology (GO) and the Enzyme Commission (EC) system both provide isa-hierarchies of enzyme functions. On the face of it GO and EC should be invaluable resources for supporting automation within Functional Genomics, which aims at predicting the function of unassigned enzymes from the genome projects. However, neither GO nor EC can be directly used for this purpose since the classes have only a natural language description. In this paper we make an initial attempt at machine learning EC classes for the purpose of enzyme function prediction in terms of biochemical reaction descriptions found in the LIGAND database. To our knowledge this is the first attempt to do so. In our experiments we learn descriptions for a small set of EC classes including Oxireductase and Phosphotransferase. Predictive accuracy are provided for all learned classes. In further work we hope to complete the learning of enzyme classes and integrate the learned models with metabolic network descriptions to support “gap-filling” in the present understanding of metabolism.