Success in association mapping of disease genes depends on knowledge of Linkage Disequilibrium (LD) structure in candidate regions. An LD map characterising such structures is constructed by making use of the Malecot model which describes the decline of LD with physical distance based on pairwise measures of association between SNPs. The HapMap project provides a valuable resource that can be used to construct genome-wide LD maps. However, the millions of SNPs in the HapMap data pose a heavy computational challenge. This difficulty can be resolved by excluding the very dist'nt SNP pairs without losing map quality. Modern computational technology with parallel processing can be used to speed up the process of map construction. A composite likelihood approach employing LD maps for association mapping has successfully localised several causal variants. An application to Rheumatoid Arthritis (RA) is described here. This approach, utilising the genome-wide LD map, is very suitable for genome-wide association studies.