dlmap(object, phename, baseModel, algorithm=c("asreml", "lme"), fixed = NULL,
random = NULL, rcov = NULL, sparse = NULL, pedigree,
seed = 1, maxit=60, n.perm = 0, multtest=c("holm", "bon"),
alpha = 0.05, filestem = "dl", ...)
"plot"(x, chr, max.dist, qcol="light blue", mcol="red", pcol="purple", marker.names=FALSE, ...)
profileplot(object, ...)
"profileplot"(object, chr, marker.names=TRUE, QTLpos=TRUE, pch=20, ...)
"summary"(object, ...)dlmapfixed, random,
sparse, and rcov separately. If a base model has already been
fit in asreml-R for the phenotypic variation, this can be input directlyidname variable. Rownames and colnames for the kinship matrix must match the idname variable. n.perm=0 (default) the Holm
correction is used. lmedlmappar for colour options)par for colour options)par for colour options)TRUE then marker names are plotted along the top of the profileplot. Defaults to TRUE. For plot function, if TRUE then flanking marker names are highlighted. Defaults to FALSETRUE then QTL positions are indicated with vertical lines in profileplot. Defaults to TRUEasreml for final model
containing all terms in the base model, as well as effects for every QTL
detected at the appropriate locations. No random effects for markers are fitalgorithm.
algorithm="asreml" provides a much more general implementation of the
DLMapping algorithm and is the preferred method of analysis.
algorithm="lme" is more restricted in its capabilities, in
that it cannot model random effects or covariance structure, cannot handle
more than 200 markers, and only allows for a single phenotypic observation
per genotype. Also, permutation has not been implemented for this function
because it is very slow. However, this version will fit the basic
algorithm and is useful should a license for ASReml not be available. In studies where the number of genetic markers is much larger than the number of phenotyped individuals, we reduce the dimension of the analysis to the number of genetic lines used in the analysis multiplied by the number of chromosomes in the genetic map. This is done in a similar manner to wgaim, with thanks to Julian Taylor and Ari Verbyla for the suggestion. The transformation of the genetic data reduces the time for computational analysis for high-dimensional data and is particularly useful in association analysis.
This version of wgaim allows high dimensional marker information to be analysed. A simple transformation of the collated high dimensional marker set shows that it may be reduced to the number of genetic lines used in the analysis. This transformation is internal to the wgaim.asreml call and users can now expect a considerably large acceleration in the performance of wgaim.
In the asreml version, there are two options for specifying the model for
phenotypic variation.
The individual model components can either be input directly as they would be
in an ASReml call, or a previous model (baseModel) output from ASReml
can be input and the components will be retrieved from it. The latter
formulation may be useful if prior phenotypic modelling has taken place. Note
that in either case, variables appearing in the rcov statement must be
ordered appropriately in the dataset. For example, if
rcov=~ar1(Column):ar1(Row) the data must be sorted as
Row within Column.
Missing values in asreml are replaced with zeros, so it is important
to centre the covariate in question. This is done for all genotypes when
algorithm="asreml". Thus individuals with phenotypic but not
genotypic data, which play important roles in field trials, may be included
safely. When algorithm="lme" these individuals cannot be included, so the
default behavior is to omit observations with missing values.
It is recommended that n.perm be set to 0 for initial exploratory
analysis, as the permutation analysis may be lengthy. The Holm
correction is used to adjust for the number of chromosomes under
consideration at each detection stage. While this is a conservative
measure it seems to perform well in practice.
Two files are output with names set by the argument filestem, which
has a default value of "dl". The
file "filestem.trace" contains ASReml licensing and likelihood convergence
output which otherwise would be dumped to the screen and possibly obscure
other messages. Errors, warnings and other messages will still appear on the
screen. Some warnings which appear may be passed through from an ASReml call
and output on exit. These may generally be ignored. This file is not created
if algorithm="lme" is used.
The file "filestem.det.log" is a record of iterations in the detection stage.
For each iteration the REMLRT testing for genetic variation on each chromosome
is output, along with adjusted p-values, genomewide threshold and markers
selected as fixed effects. The p-values are corrected for the
number of chromosomes tested either by the Holm correction or by
permutation. If the number of permutations (n.perm) is greater than
0, then for the Xth iteration an additional file "filestem.permX" will be
created which contains the test statistics for the permuted datasets.
See the accompanying vignette for an example of how to interpret the ".det.log" file.
If the type of cross is not "other", the plotting function plots the genetic linkage map for a selection of chromosomes. Indicates marker
locations, marker names, and detected QTL positions and associated flanking
markers obtained from a dlmap fit. This function relies upon link.map.cross, which was written by Julian
Taylor for the wgaim package. It is built upon here by adding QTL
regions and estimated positions to the map.
The function plot.dlmap provides a neat visual display of
chromosomes. If no QTL are detected, only the linkage map will be plotted;
otherwise detected QTL will be placed at their estimated positions
and the intervals around them (and flanking markers) will be highlighted.
If a subset of chromosomes are plotted and detected QTLs exist outside that
subset a warning will be given that QTLs have been omitted from the display.
The arguments mcol, qcol and pcol have been added for
personal colour highlighting the flanking markers, QTL regions and QTL
positions respectively. The procedure may also be given the usual col
argument which will be passed on to the other markers.
In order to ensure that all marker names are displayed without vertical overlap, the default value of the "cex" parameter passed to "text" should be manipulated. For large maps with many chromosomes, marker names and adjacent chromosomes will overlap horizontally. In this case it is suggested that the user horizontally maximize the plotting window to remove overlap, or subset the chromosomes displayed.
The profileplot function plots the Wald statistic profile for each chromosome with detected QTL on first
interval mapping scan. Indicates marker locations, marker names, and detected
QTL positions obtained from a dlmap fit. It provides a neat visual display of
the Wald profile for chromosomes with detected QTL. If no QTL are detected,
nothing will be plotted. Otherwise, the Wald profile will be plotted by cM
position of points on the interval mapping grid. Marker names will be
displayed at the appropriate positions along the top of the plot. Vertical
lines will mark the position of detected QTL.
The summary function outputs a summary of a dlmap object and detected QTL. It primarily prints the summary table
computed from dlmap. This includes the
chromosome QTL are detected on, estimated positions, flanking markers, QTL
effects and standard deviations, Z-ratio and p-value.
B. Emma Huang, Rohan Shah, Andrew W. George (2012). dlmap: An R Package for Mixed Model QTL and Association Analysis. Journal of Statistical Software 50(6): 1-22. URL http://www.jstatsoft.org/v50/i06/.
dlcross ## Not run:
# data(BSdat)
# data(BSphe)
#
# # Convert cross object to DL Mapping format
# dl.in1 <- dlcross(format="rqtl", genobj=BSdat, idname="ID", fixpos=1)
#
# # Analyze data
# BSdl <- dlmap(object=dl.in1, algorithm="lme", phename="phenotype", filestem="BS")
#
# plot(BSdl)
#
# # With additional phenotypic data
# dl.in2 <- dlcross(format="rqtl", genobj=BSdat, pheobj=BSphe, idname="ID", step=5)
# BSph <- dlmap(object=dl.in2, algorithm="asreml", phename="phenotype", env=TRUE, random=~Block)
#
# profileplot(BSph)
# summary(BSph)
# ## End(Not run)
Run the code above in your browser using DataLab