When interpreting results for pedigree analysis, it is extremely helpful to know how often a similar result might arise by chance. For example, in a linkage analysis it may be helpful to know how many peaks of similar height are expected conditional on the set of phenotypes being analysed and the available marker map. When investigating suspicious genotypes, it is important to characterize the false-positive rate for error detection procedures.

MERLIN has the ability to perform gene dropping simulations which replace input data with simulated chromosomes conditional on family structure and actual marker spacings and allele frequencies, as well as missing data patterns. The procedure for generating simulated data is described in the reference section.

For this example, we will use a data set from the examples subdirectory of the MERLIN distribution as input. You can also find the example data in the download page.

Estimating false positive rates for error detection

In the error detection tutorial, we identified 7 pairs of unlikely genotypes in a 20 marker, 5-cM scan, of 200 sib-pairs, corresponding to 8,000 total genotypes. The data is organized into three files, a pedigree file summarizing genotypes and relationships (error.ped), a data file describing the contents of the pedigree (error.dat) and map file providing marker locations (error.map).

To review a descriptive summary of the dataset, you could run pedstats:

prompt> pedstats -d error.dat -p error.ped

To review the original set of unlikely genotypes, you could use MERLIN's automated error analysis:

prompt> merlin -d error.dat -m error.map -p error.ped --error

To estimate false positive rates, we will request that MERLIN analyse a simulated data set with identical allele frequencies and marker spacing by using the --simulate command line option. Try it out!

prompt> merlin -d error.dat -m error.map -p error.ped --error --simulate

You should first see the MERLIN start-up screen and summary of selected options. Note that the options --error and --simulate are selected. Note also that the current random seed is 123456. This seed indicates which simulated replicate will be used, and selecting a different seed produces an alternative simulated data set.

MERLIN 0.8.4 - (c) 2000-2001 Goncalo Abecasis

The following parameters are in effect:
                     Data File :       error.dat (-dname)
                 Pedigree File :       error.ped (-pname)
            Missing Value Code :         -99.999 (-xname)
                      Map File :         asp.map (-mname)
            Allele Frequencies : ALL INDIVIDUALS (-f[a|e|f|file])
            Steps Per Interval :               0 (-i9999)
                   Random Seed :          123456 (-r9999)

Data Analysis Options
         General : --error [ON], --ibd, --kinship, --information
         Linkage : --npl, --pairs, --qtl, --deviates, --vc
     Haplotyping : --best, --sample, --all, --founders
   Recombination : --zero, --one, --two, --three, --singlepoint
          Limits : --bits [24], --megabytes
          Output : --quiet, --markerNames
      Simulation : --simulate [ON], --save
      Additional : --simwalk2, --matrices, --swap

This start-up screen should be followed by an error detection analysis for the replicate, which should indicate a single pair of unlikely genotypes:

Family:    38 - Founders: 2  - Descendants: 2  - Bits: 2
  MRK6 genotype for individual 3 is unlikely [0.021855]
  MRK6 genotype for individual 4 is unlikely [0.021855]

NOTE: In many newer versions of MERLIN, you may not find any unlikely genotypes in the replicate produced with the default seed. This is not a problem, and merely reflects the low false positive rate of the procedure. Continue reading to learn about how to use a different seed...

So MERLIN flags a single pair of unlikely genotypes in this particular replicate... Is this typical of other replicates? There are two ways to investigate the issue further.

One option is to generate additional replicates, one at a time, by repeating the above procedure with a different random seed. To do this, you will need to set the -r command line option. The following command repeats the previous analysis but sets the random seed to 1234, thus generating a different set of simulated data:

prompt> merlin -d error.dat -m error.map -p error.ped --error --simul -r 1234

Another option is to request that MERLIN loop through the simulation procedure multiple times. This option is available through the --reruns command line option in newer versions of MERLIN. To analyse 20 simulated datasets, try:

prompt> merlin -d error.dat -m error.map -p error.ped --error --simul --reruns 20

In either way, it is straight-forward to repeat any MERLIN analysis for simulated chromosomes and estimate false-positive rates for error detection or linkage analysis (note that MERLIN does not change input phenotypes and disease status when conducting simulations).

Although we focused on simulating data under the null hypothesis (that is on simulating random genotypes that are independent of the phenotype and genotype data), Merlin can also simulate quantitative trait loci associated with a specific simulated phenotype. The procedure for these simulations under the alternative hypothesis is sketched out in the reference section.

Now that you have seen how to generate simulated replicates, you could proceed to haplotype analysis or ibd estimation. If you haven't already done so, you could try the linkage or error detection tutorials.