1Log-linear Modelling of SNP Haplotype Blocks

Algorithms for defining the presence and limits of haplotype blocks have been presented previously, using measures of linkage disequilibrium. In the current manuscript, a new log-linear framework is proposed for the identification of blocks. Our approach allows a reparameterisation and formalisation of existing methods, including a means to statistically test, via the likelihood ratio test, for the presence of block-like structure. Our method was applied to a data set of 76 SNP markers spanning a genomic interval of length 2.7Mb. Obvious blocks were verified by our approach. In addition, evidence for sub-structure within blocks was also detected, suggesting that blocks, if they exist, may be far more complex than previously assumed. It is hoped that our approach will provide the basis for a formal statistical evaluation of blocks and will facilitate comparisons among data sets