Class McLink
java.lang.Object
McLink
public class McLink
- extends java.lang.Object
This program implements the methods of
Thomas, A., Gutin, A., Abkevich, V. & Bansal, A. (2000), Multipoint linkage
analysis by blocked Gibbs sampling, Statistics and Computing
10, 259-269
in order to calculate linkage statistics using Markov chain Monte Carlo methods.
Usage java McLink input.par input.ped [n_simulations] [-v/-t] [ld.model]
where
- input.par is a LINKAGE parameter file
- input.ped is a LINKAGE pedigree file
- n_simulations is the number of MCMC samples to use in the calculations.
The default is 1000.
- If -v option is specified lod scores are output after each iteration which
may be useful in tracking the performance of the sampler. If -t is
specified there is no intermediate output. -t is the default.
- ld.model is an optional file specifying a graphical model for linkage disequilibrium.
Such a file can be produced using HapGraph if a population sample of genotyped
individuals is available.
The assumption is that McLink will be used without an LD model when the data are
micro satellite genotypes and with an LD model when the data are SNPs, hence, there
are differences in what is assumed and what is possible.
Run without a specified linkage disequilibrium model:
- primarily for use with micro satellite markers.
- loci are assumed to be in linkage equilibrium.
- any genotyping errors that lead to Mendelian inheritance
inconsitencies will crash the program. Check the data with GMCheck first.
- space of possible genotype states at each locus is downcoded to speed up the
computations.
- the sampler used is an extended LM-sampler which has better mixing properties.
Run with a specified linkage disequilibrium model:
- primarily for use with SNP genoypes.
- specified LD model is assumed.
- genotyping error is modelled assuming an error rate of 1%
and that the probability of error is independent of the actual
genotype. The assumption here is that SNP loci are individially relatively
uninformative and there are too many of them to make error checking and
correction feasible.
- no downcoding of the state space is done.
- the sampler used is an L-sampler since the LM-sampler is not possible
with an LD model. The mixing properties are probably not great.
In the output, for each kindred in the pedigree file there is a matrix
of TLOD, or complex lod score statistics, with a row for each locus and
a column for each value of the recombination fraction in
{0.0, 0.01, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.99, 1}.
|
Method Summary |
static void |
main(java.lang.String[] args)
|
| Methods inherited from class java.lang.Object |
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
McLink
public McLink()
main
public static void main(java.lang.String[] args)