In this preliminary work, we investigate a method to model linkage disequilibrium among SNPs (Single Nucleotide Polymorphisms) in the genome. The genetic data such as SNPs is characterized by a typical block-like structure along the genome. Graphical models such as Bayesian networks can provide a fine and biologically relevant modeling of dependencies for both haplotypical and genotypical SNP data. We applied a MWST-based algorithm (Maximum Weighted Spanning Tree) to construct a Bayesian network, relying on the underlying local dependencies.