Advances in Statistical Human Genetics Over the Last 25 Years
Twenty-six years ago restriction fragment length polymorphisms, and a decade later microsatellites, were proposed as markers for whole genome linkage analysis, leading to the development of multipoint linkage analysis, both model-based and model-free. Whereas the Elston-Stewart algorithm was already available for model-based analysis, it was not well-suited to multipoint linkage analysis until algorithmic improvements were developed, and even then it was limited to relatively small numbers of markers. The Lander-Green algorithm and its extensions made multipoint linkage analysis feasible, whatever the number of markers - its main limitations being the size of the pedigrees it can feasibly handle and the requirement that the markers all be in linkage equilibrium. Since then bead and chip technology for single nucleotide polymorphisms, and the success of the HapMap project, have made linkage disequilibrium mapping feasible. This has led to the development of multiple new methods of analysis, as well as novel designs, for both linkage analysis and association analysis. Of particular concern has been the possibility of association studies being confounded by population stratification and the multiplicity of tests now possible. This has motivated the development of further novel methods, of both design and analysis. These various designs and analytical methods will be reviewed and critically compared with respect to their robustness and power for finding genomic locations and their efficiency for estimating genetic effects, especially in the case of genes underlying complex diseases.