crimap documentation (version 2.4)
7. Changes incorporated in version 2.2 - 2.4
The main change in version 2.2 is the incorporation of a more
efficient "switch algebra" algorithm which in many cases (particularly
for large pedigrees) substantially reduces memory requirements
and running time. Thus for many runs, even with a relatively
large number of loci, the default memory request of 3 MB
(for nb_our_alloc in the
.par file) is overly generous. Version
2.3 incorporates a better switch algebra algorithm which further
reduces memory and running time. Additional changes have been
made which should make this version compatible with most C compilers.
Version 2.4 incorporates the following changes:
- Haplotyping: it is now possible to constrain the loci in a haplotyped
system to be at 0 distance from each other for all likelihood
calculations. In addition, haplotyping can be used with all
of the options; in particular,
twopoint LOD tables can be generated
for pairs of haplotyped systems. Haplotyped systems are
now specified in the
.par file (and there is no longer a .hap
file).
- Fixed distances: it is now possible to fix the distance
between a pair of adjacent loci, for the purpose of likelihood
calculations, with the options fixed
and chrompic.
- The .ord
file can optionally be used to prescreen locus orders generated
by all and flipsn, to reduce the number of orders for
which likelihoods need be calculated. (Previously the
.ord file
was only used by the mapbuilding options). The parameter
use_ord_file in the
.par file controls the use of the
.ord file
with these options.
- twopoint: it is now possible to divide
the loci into two different groups, and generate
LOD tables only
for pairs consisting of one locus from each group (this makes it
possible, for example, to conveniently generate
LOD tables
between "new" loci and "old" loci). The parameter PUK_LIKE_TOL
is now used as a
LOD cutoff: tables are displayed only for locus
pairs whose
LOD exceeds PUK_LIKE_TOL.
- chrompic: chromosomes
are now labelled by individual ID # (not position within
.gen
file); phase designations are now always consistent across pedigrees
(not merely across chromosomes) even in the phase unknown
case; the relative probabilities for the best phase choice and
the next most likely alternative are given; the phase designations
now distinguish phase known from phase unknown.
- a new
option, merge, makes it easy to merge two
.gen files having overlapping
sets of families and/or loci. Other changes in v. 2.4,
include: family IDs are now allowed to be arbitrary character
strings (not merely integers as in previous versions); the format
of the
.par file has been changed to make it easier to use; the
orders displayed by all are sorted by decreasing log10 likelihood;
output of flipsn is easier to read (the "flipped" loci in each
order are highlighted).
up: table of contents
previous section: 6. technical notes
next section: 8. bibliography