Extending Logic Programming for Life Sciences Applications

The volume of bioinformatics data produced by research laboratories worldwide is increasing at an astonishing rate turning the need to adequately catalogue, represent and index the vast amounts of produced data sources into a pressing challenge. Semantic Web technologies have achieved signi cant progress towards the federation of biochemical information via the de nition and use of domain vocabularies with formal semantics, also known as ontologies. OWL [4], a family of logic-based knowledge representation formalisms, has played a pivotal role in the advent of Semantic Web due to its signi cant ability to reason over ontologies by means of logical inference. A core reasoning task in biochemical domains is classi cation; with the help of performant OWL reasoners life scientists can employ OWL to drive fast, automatic and repeatable classi cation processes. As a consequence, OWL bioand chemo-ontologies are widely used for the modelling of life sciences knowledge. A prominent Semantic Web ontology is ChEBI, an open-access dictionary that provides taxonomical information for molecular entities with numerous applications such as drug discovery and study of disease pathways. Inspite of ChEBI being available in OWL format, classifying ChEBI with an OWL reasoner produces only few of the desired subsumptions because OWL is not expressive enough to capture the structure and the properties of chemical entities. For instance, the tree-model property of OWL prevents one from faithfully representing cyclic molecules: OWL axioms can state that butane molecules have four carbon atoms, but they cannot state that the four atoms in a cyclobutane molecule are arranged in a ring. Additionally, due to the rst-order logic semantics of OWL it is di cult to describe classes based on the absence of certain characteristics: if an inorganic molecule is de ned as `a molecule not containing a carbon atom', then in order to derive that water is inorganic one needs to also specify that water contains at most three atoms and that both hydrogens and oxygens are disjoint from carbons. As a consequence of these inadequacies, ChEBI is manually curated by human experts who determine the chemical classes of new molecular entries. Currently, ChEBI describes nearly 30,000 entities and grows at a rate of 3,500 entities per year. Given the size of other publicly available chemical databases with millions of molecules, there is clearly a strong potential for ChEBI to expand by speeding up the curating tasks through automation of chemical classi cation. Towards that direction, we developed a theoretical and practical framework for the representation of graph-shaped objects, with a particular application to chemistry. Namely,