The spatial dimension in biological data mining

Among its numerous applications, data mining plays an increasingly important role in epidemiology. In particular, it allows processing the steadily increasing volume of genomic data and helps identifying genetic risk factors. Despite ongoing progress, the mining methods currently manufactured for exploring such data still stumble over their very characteristic features and in particular their considerable complexity and diversity. Genomic data range from DNA sequences and single nucleotide polymorphisms (SNPs) to gene and protein expression levels and protein-protein interaction patterns, and further encompass structural and functional genome annotation. Accordingly, various types of data are generally treated independently and patterns emerging from any set of analyses are stitched together to form a biological answer or to generate new hypotheses.Occasionally, such patterns are projected onto a geographical map, superimposed to migration patterns or correlated to environmental factors, placing crude numeric information into a spatio-temporal perspective [reviewed in [1]]. Integrating spatial, environmental and genetic data into models of geographic disease etiology (ecogeographic genetic epidemiology) has recently been proposed as a new interdisciplinary pathway to understand the distribution and the determinants of diseases [1]. The Geographic Information Systems (GIS) used to integrate these multiple layers of information is a set of powerful hardware and software for inputting, managing, displaying and analyzing geographically referenced information. GIS have relatively recently been recognized as a useful tool for biomedical research, and in particular for visualizing cancer distributions and estimating the contribution of various environmental risk factors to cancer prevalence [reviewed in [1]]. Accordingly, the American National Cancer Institute http://gis.cancer.gov/ webcite, with the Long Island Breast Cancer Study Project for instance http://li-gis.cancer