EVALUATION OF TWO ALGORITHMS FOR THE CONSTRUCTION OF GENETIC LINKAGE MAPS

Department Plant Breeding PO Box 386 6700 AJ Wageningen The Netherlands

Most algorithms for genetic map construction use pair-wise marker distances. The marker order is obtained by minimizing the differences between recombination frequencies from the pair-wise data and calculated frequencies in the map. For linkage groups containing more than 200 markers this method appears to be very time-consuming. We propose a novel procedure in which maps are built directly from the raw data. First, data redundancy is removed by eliminating cosegregating markers. Next, a preliminary order is obtained by using the total number of recombinations as a criterion for the best position. Finally, a search for the minimum number of recombinations is performed by systematically swapping a moving window of markers of increasing size. To compare the performances of both methods, maps were constructed from simulated data sets varying in population sizes, number of loci, percentage missing data and error rate, after which rank correlation coefficients between the true and calculated orders were assessed. Results of both methods were hardly influenced by missing data, but the introduction of errors in the data caused a substantial decrease in performance. The new method appeared to be less sensitive to errors than the pair-wise method. The new method also excels in calculation speed, reducing the 30 hours processor time needed by the pair-wise method to calculate a large map to merely one hour. This improved speed and performance make the new method attractive especially for high-density mapping projects.