Extracting structured and compact knowledge from massive amounts of experimental data is a major challenge for bioinformatics. Numbers of algorithms and tools have been developed to process experimental information from high-throughput experiments. Current approaches to the processing of high-throughput experiment data consist of two stages: primary and secondary. The primary processing stage involves normalization, conversion and filtering of the raw experimental data. The secondary stage is set to structure and condense the results of large-scale experiments making them assessable by a human expert.
In this work we concentrate on enrichment analysis algorithms used at the secondary stage of data processing. We propose a novel approach to the bioinformatics enrichment tools which allows us to discover and analyze statistical fluctuations of complex combinations of annotation terms and annotation term relations. The approach is based on the application of Inductive Logic Programming (ILP) to discover potentially enriched annotation concepts. The annotation information is represented as first order logic statements. The discovery process is formulated as an ILP classification problem, where the resulting set of rules represents detected annotation concepts. Statistical analysis is consequently used to establish enrichment scores of annotation terms and concepts. Based on the scores and given output threshold annotation terms and concepts are sorted and filtered.