Supporting Discovery in Biomedicine by Association Rule Mining of
Bibliographic Databases

Dimitar Hristovski, Borut Peterlin, Saso Dzeroski
University of Ljubljana

Positional cloning approach has proved very successful in cloning genes for
Mendelian genetic human diseases. However, gene identification rarely implies
understanding of pathophysiology of a genetic disease and consequently the
rationale for therapeutic strategies. Moreover, knowing the entire human genome
sequence requires novel methodological approaches in the analysis of genetic
disease. In this paper we describe an interactive discovery support system (DSS)
for the field of biomedicine in general and human genetics in particular. The
intended users of the system are researchers in biomedicine. The goal of the
system is to discover new, potentially meaningful relations between a given
starting concept of interest and other concepts that have not been published in
the biomedical literature yet. The main idea is to first find all the concepts Y
related to the starting concept X (e.g. if X is a disease then Y can be
pathological functions, symptoms, etc.). Then all the concepts Z related to Y
are found (e.g. if Y is a pathological function, Z can be a molecule,
structurally or functionally, related to the pathophisiology). As the last step
we check whether X and Z appear together in the medical literature. If they do
not appear together, we have discovered a potentially new relation between X and
Z. This relation should be confirmed or rejected using human judgment,
laboratory methods or clinical investigations, depending on the nature of X and
Z. The known relations between the medical concepts come from the Medline
bibliographic database. The concepts are drawn from the MeSH (Medical Subject
Headings) controlled dictionary and thesaurus, which is used for indexing in the
Medline database. We use a data mining technique called ?association rules? for
discovering relationships between medical concepts. Our discovery support system
is interactive, i.e. that the user of the system can interactively guide the
discovery process by selecting concepts and relations of interest. The system
provides the possibility to show the Medline documents relevant to the concepts
of interest and also to show the related proteins and nucleotides. We used the
DSS to analyze Incontinentia pigmenti (IP), a monogenic genodermatosis, the gene
of which has recently been identified via the positional cloning approach
(Nature 2000;405:466-71). We were interested whether the gene could have been
predicted as a gene candidate by the DSS. We succeeded in identifying the NEMO
gene as the gene candidate and in retrieving its cDNA sequence (available since
1998). Moreover, the DSS provided some potentially useful data for understanding
the pathogenesis of disease. It has to be stressed that efficient use of DSS is
largely driven by the scientist. We conclude that the DSS is a useful tool
complementary to the already existing bioinformatic tools in the field of human
genetics.