Running an analysis

QTL Cartographer was originally written for use on Unix workstations. The main documentation refers to a series of programs that can be used for a QTL analysis. If you get serious about QTL analyses, you'll need to understand each of those programs thoroughly. The design of QTL Cartographer reflects the Unix philosophy. Instead of writing one, big, monolithic program that does everything. Write a series of small tools that work well together, and allow users to work with them individually as they see fit. That's great for complex analyses, but it makes learning the package somewhat more difficult. For our purposes, we'll stick with the simpler interface provided by WinQTLCart. After loading the data with File->Open and hitting the Verify button, your screen should look something like Figure 1.

Figure 1: Screenshot of WinQTLCart after the data in Maize.mcd has been loaded.

You could modify information about the traits, the genetic map, or the crosses using the buttons if you chose to, but we'll assume that everything is as it should be so that you don't have to worry about that. If you now hit the DrawChr button, you'll get a nice graphic showing you a linkage map of the genetic markers you're using (Figure 2).

Figure 2: Map of genetic markers for the sample data set used in this analysis.

Once you've made it this far, running the analysis is as simple as pulling down the Method menu and selecting the method you want to use. We don't have time to discuss the differences among the methods in detail, but I'll briefly summarize the methods here:

Single Marker Analysis

Select Single Marker Analysis and press ``Go'' (or select Method->Single Marker Analysis from the menu).

• If you push the View info... button in the Single Marker Analysis box, you'll get a linear regression analysis of the relationship between phenotype and marker genotype for each marker individually. This analysis tells us that there's a significant positive relationship between genotype and phenotype for the first three markers.²

• If you push the View info... button in the Statistical Summary box, you'll get summary statistics on the pattern of trait variation in the mapping population and on the pattern of segregation at the marker loci, i.e., whether they follow Mendelian expectations. The values should follow a $\chi^2$ distribution with 1 degree of freedom. In this case, the genotype proportions in our mapping population all appear to be consistent with Mendelian expectations.

Interval mapping

When you select this menu item the analysis performed is simple interval mapping of the type I've already described. One slight complication is that because you're doing a lot of statistical tests when doing a QTL analysis, you have to take account of that fact in choosing a threshold value of the likelihood ratio statistic for declaring that you've found a QTL. You can accept the default value, put in one of your own choosing, or select one through permutations (which will take the longest, but should produce the most reliable choice). After you push the OK button, you'll see the program counting down from the number of permutations you asked for to zero.

The other parameter you may want to change is the Walk speed. That's the parameter that determines the interval along the map at which QTL calculations are done. If you have a very dense map, you can set the interval to be quite small, and you'll have a much more precise idea of where any QTLs you locate may be, but it will take the program much longer to do the calculations. We'll leave the walk speed at the default 2cm for this example.

Once the permutations have finished, WinQTLCart will automatically enter the new threshold value, and you're ready to look for a QTL. Hit the Start button, and you'll soon see something like Figure 3

Figure 3: Interval mapping results for the sample data. I turned off the background using Settings->Show Colorful Background option to make the background ywhite.

This figure suggests that a QTL is present at about 6cm from the left end of the chromosome. Finding the corresponding line in the output (position 0.0601) we see that the additive effect of the $A$ allele at this locus is estimated to be 1.16, the dominance deviation (the extent to which the heterozygote departs from intermediacy) is 0.0388, and that this QTL accounts for about 22% of the variance in the trait.³

Composite interval mapping

The options available under composite interval mapping are very similar to those for interval mapping. That's because the underlying statistical model is very similar. In fact the only difference is the CIM is attempting to statistically control for the genotype at markers other than those immediately flanking the candidate QTL. The results are in Figure 4. They look pretty similar, but notice that the peak is at 0cm and the one at about 47cm barely reaches the threshold.

Figure 4: Results of a composite interval mapping analysis of the sample data.

Multiple interval mapping

Multiple interval mapping is a still more sophisticated method of mapping. It allows you to identify more than one QTL and to refine your analyses as you go along. One nice feature is that it puts a nice summary of the results up in the window. The data we've been using give us a QTL at 3cm, with an additive effect of 1.15, and a dominance deviation of 0.069. Running the summary statistic report, we find (again) that this QTL explains about 19% of the phenotypic variance.

Bayesian interval mapping

⁴ Although Bayesian interval mapping appears on the menu, and an analysis will run, I haven't had time to figure out how to interpret the results yet, so we won't talk about it.