oneMarkerDistribution() now fully supports pedigrees
with multiple components.
oneMarkerDistribution() has a new argument
output taking values "array" (default, as
before), "table", or "sparse". Both
"table" and "sparse" produce data frames where
each row corresponds to a genotype combination. The
"sparse" format only includes combinations with non-zero
probability.
In oneMarkerDistribution(), the argument
partialmarker has been renamed to marker. The
old name still works as an alias, but will be removed in a future
version. While the old argument had no default value, the new defaults
to the first attached marker. This simplifies the call in many cases,
for example
singleton("A") |> addMarker() |> oneMarkerDistribution("A").
In twoMarkerDistribution(), the arguments
partialmarker1 and partialmarker2 have been
renamed to marker1 and marker2,
respectively.
Fixed bug affecting likelihood calculations in pedigrees with
partial genotypes (e.g. "1/-") in founders.
The function allGenotypes() is ~2-4 times faster due
to a better implementation.
Updated dependencies: pedtools v2.6.0, pedmut v0.7.0.
Fixed rare bug in the peeling algorithm manifesting with reversed peeling order.
pedtools v2.2.0 is now required.
Fix bug in likelihood() affecting singleton with
partial genotype (e.g. “1/-”).
Minor optimisations in reduceAlleles() and
startdata_M().
More efficient oneMarkerDistribution() and
twoMarkerDistribution() in cases without mutation
models.
README updates.
The version implements several improvements in the peeling algorithm, mostly invisible to end users (except that likelihoods calculations will be noticeably faster, and use much less memory, in many cases).
Less memory footprint for untyped founders with only 1 child: Sample alleles directly, avoid complete set of genotypes.
Avoid costly 3-dimensional arrays in peeling calculations.
More efficient genotype elimination in likelihood calculations
involving markers without mutation modelling. As a result, this is now
enforced (rather than optional) everywhere, and the
eliminate argument is deprecated.
Added citation info.
Mutation models are now supported in likelihood2(),
providing likelihood calculations for pairwise linked markers.
New functions haldane() and
kosambi().
Update README.
Update package doc.
Include license (GPL >= 2).
reduceAlleles() is faster in some cases, due to the
new pedmut::lumpedModel().
A few minor speedups and code improvements.
merlin() gains argument checkpath. Set
this to FALSE to avoid redundant checks of MERLIN availability.
likelihoodMerlin() gains argument
perChrom, used to parse the chromosome-wise likelihoods
from the MERLIN output.
lumpAlleles() gains argument always. By
default lumping is skipped for markers where all individuals are
genotyped, but this can cause problems e.g. with MERLIN which operates
with an upper limit of alleles.
Some efforts are done to check for (and warn about) underflow/overflow in MERLIN results.
checkMerlin() by default
checks that both merlin and minx are available
on the system, and that they both come from the latest MERLIN
version.pedprobr now depends on R 4.1 and
pedtools 1.1.0.
The deprecated argument loop_breakers has been
completely removed (renamed to loopBreakers)
likelihood.ped() gains a new argument
lump, which activates allele lumping. This is TRUE by
default, and should remain so in most cases.
Using theta correction in pedigrees with inbred founders now gives an error.
The internal peelingProcess() now checks for
unbroken loops in the pedigree. (Of interest for developers
mainly.)
When calling MERLIN, the window-centric and unnecessary
.exe extensions has been removed.
Many examples have been rewritten in simpler code, taking
advantage of pedtools::addMarker and the pipe
|>.
setMutationModel(..., model = NULL).setMutationModel() affecting lists of
multiple pedigrees.likelihood() has been refactored, moving the treatment
of (two) linked markers to a separate function,
likelihood2().Theta-correction is implemented in likelihood calculations,
through the argument theta of
likelihood().
merlin() and likelihoodMerlin() have
been overhauled.
A new function checkMerlin() checks if MERLIN is
installed and available.
merlin() now has an argument linkageMap
facilitating analysis of linked markers.
likelihoodMerlin() gains arguments rho
and logbase.
New function lumpAlleles().
The likelihood() function now handles multiple
marker inputs. For example, the call likelihood(x, 1:2)
results in a vector of length 2 with the likelihoods of the first two
markers attached to x.
Recombination parameter theta is renamed to
rho everywhere, to align with other pedsuite
packages.
Many input checks have been added, giving more useful error messages.
The data structure used in likelihood calculations is simplified and more efficient.
Likelihood computation of two linked markers (and therefore also
twoMarkerDistribution()) is much faster now.
New general MERLIN wrapper, merlin().
Both likelihoodMerlin() and the new
merlin() now accepts ped lists.