Tables 7.1-7.2 summarize the timings for Zmapqtl to do interval mapping [Lander and BotsteinLander and Botstein1989] and composite interval mapping [ZengZeng1993,ZengZeng1994] on various computing platforms under different operating systems. All timings were done in the winter of 1999. The simulated data set has been used previously [ZengZeng1994] and consists of a genetic linkage map that has four chromosomes with 16 markers on each chromosome. The markers are evenly spaced at 10 cM and the simulated data has one trait. The entire genome was scanned at a walking speed of 2 cM. The programs were run in automatic mode, with no recourse to the interactive menus. They indicate the amount of time to read in the data, perform the analysis and write the output.
Table 7.1 summarizes the timings for interval mapping.
Table 7.2 summarizes timings for composite interval mapping. The model for analysis was Model 6 with a window size set to 10.0 cM and using up to 5 markers to control for the genetic background. Some of the ratios for the same machine change from interval mapping to composite interval mapping. Model 6 uses quite a lot more double precision arithmetic, and this may account for the differences.
For 
replications of a permutation test
or bootstrap, the computing time should be less than times the
values in Tables 7.1-7.2. The jackknife analysis
should be around 
times these values (where is the sample size).
|
The MImapqtl module was run on numerous systems for the mletest data set with information criterion 1 and a threshold of 0.0. Timings are presented in Table 7.3. The systems ran MacOS 8.6, Solaris 8 and Windows 2000, respectively for the Macintosh, UltraSparc and Pentium III machines. The analytic results were the same on all platforms. MImapqtl is sensitive to the number of parameters in the model it is analyzing. In this example, 10 QTL were identified. the first stages of QTL identification proceed quickly, but the program slows down significantly as more parameters are used. This is something to keep in mind if you decide to search for epistatic terms. If there are a lot of main effects, then a backward elimination method for finding epistatic interactions may take a prohibitavely long time.
|