Statistical analysis of molecular genetic data

Recent advances in molecular biology have made it possible for geneticists to collect data at the DNA sequence level, and so make observations on genes directly, instead of on characters affected by genes. This new nature of genetic data is posing new problems of statistical analyses, as is shown in this paper. After an explanation of the methodology of recombinant DNA studies, separate consideration is given to the statistical analyses of restriction-map and complete-sequence data. The initial observations from restriction-enzyme studies are of the number and sizes of the fragments produced by digesting a DNA region with the enzymes, and it is necessary to infer the location of the breakage points. Model-based analyses are then necessary to infer characteristics of the DNA sequence from the frequency of these break points. An important application of such restriction maps is the detection of human disease genes, with the most notable success to date being the chromosomal allocation of the gene for Huntington's chorea. Complete sequence data present conflicting problems of scale. The amount of information collected per individual is so large that analyses must always be done on a computer, but the number of individuals studied in a population is too low to allow the magnitudes of between- and within-population variation to be assessed. Statistical problems also include the detection and testing of sequence patterns and the comparison of sequences. Any analyses are going to have to incorporate the observed high levels of association between adjacent sequence members. While national data bases now contain about three million sequence elements, very little is known about the error rate in the generation or recording of these sequences.