Recent_Changes_v16.txt

 New Features in Mendel Version 16

In version 16 of Mendel, the expiration date has been removed.
Also, two new analysis options have been added, both quantitative
versions of the Maternal-Fetal Genotype (MFG) incompatibility test.
These are QMFG LRT (Analysis Option 30) that performs a Likelihood
Ratio Test (LRT), and QMFG Score (Analysis Option 31) that uses the
Score test instead of the LRT. The score test algorithm is much more
efficient than the LRT, while the LRT is slightly more accurate.
The QMFG Score Option can easily handle dense genome-wide SNP data sets.
(There was no public version 15 of Mendel.)




 New Features in Mendel Version 14

The main change in this version of Mendel is faster computation in many
of the analysis options, dramatically faster for GWAS (Analysis Option
24), Ped-GWAS (Analysis Option 29), and Kinship Estimation (Analysis
Option 10). This was predominantly obtained through parallelization. If
desired, the user can limit or turn off this parallelization by setting
the new keyword MAX_THREADS to the number of processors Mendel should
use. The default is to use all available processors.

In addition to making Ped-GWAS and Kinship Estimation much faster, they
are also now much easier to use. For example, simply setting the value
of the keyword KINSHIP_SOURCE in the control file determines if the
global kinship coefficients are estimated via (1) only the explicit
pedigree structures in the input file, (2) by comparing the SNP
genotypes of pairs of individuals within pedigrees, which is the
default, or (3) comparing the SNP genotypes of all pairs of individuals
in the data set. If the kinship coefficients are set to be derived from
SNP genotypes, the number of SNPs used can be set via the keyword
SNP_SAMPLING_INCREMENT. The default value of 5 implies 20% of the SNPs
are used; to use 100% of the SNPS, set this keyword to the value 1. The
user can also trivially set overall minimum and maximum numbers of SNPS
to use. Finally, the user can use the new keyword KINSHIP_METHOD to
choose between the Method of Moments (MoM) and the Genetic Relationship
Matrix (GRM) algorithms for converting SNP genotypes to global kinship
coefficients. On a standard laptop computer the association analysis of
1 million SNPs and 1000 related individuals often takes less than 90
seconds and requires only about 1 GB of RAM.

Many more minor additions are also included in this new version of
Mendel. For example, the new keywords MIN_MAF and MAX_MAF allow one to
easily set bounds on the minor allele frequencies of the SNPs allowed to
be included in the analysis. Also, Pedigree Trimming (Option 21) has a
new sub-option that simply combines all individuals into one pedigree,
keeping all stated relationships intact.

In the 14.x revisions of Mendel, Plink file importation was improved.
A larger change is that the default penalty function used in
penalized regression in GWAS (Analysis Option 24) was changed from
the classic lasso to the minimax concave penalty (MCP). The MCP escapes
the over shrinkage of the lasso, and thus should improve model selection
by better avoiding false positives. To force Mendel to use the lasso
penalty instead of MCP, set the new keyword LASSO_PENALTY to true.

Also, Mendel now significantly simplifies and decreases the computation
time for Ped-GWAS (Analysis Option 29) and Kinship Estimation
(Analysis Option 10). For the Ped-GWAS Option, when you list several
quantitative traits, each is analyzed separately, unless the new keyword
MULTIVARIATE_ANALYSIS is set to true. This makes performing many
univariate analyses much faster, since the data needs be initialized
and the kinship coefficients estimated only once. Also, a new shorthand
"all traits" can be used if all variables should be analyzed
or a particular predictor should be used for all traits.




 New Features in Mendel Version 13

This new version has significantly expanded several existing options and
add one entirely new analysis option. The new Ped-GWAS analysis option
provides extremely fast GWAS analysis on extended pedigrees, nuclear
families, unrelated individuals, or any combination thereof. It can
handle univariate or multivariate quantitative traits with missing data,
and allows for covariate adjustment, including correction for population
stratification. Pedigree kinships can either be explicitly provided or
automatically estimated from dense markers. SNPs can be analyzed under
codominant (additive), dominant, or recessive models.

The Ethnic Admixture analysis option has been expanded to rank markers
by their informativeness to distinguish subpopulations. This option can
now also perform very efficient principal component analysis (PCA).
These two new features can be used with either text-based or binary data
files.

The Kinship analysis option has also been expanded to binary files. It
can now quickly estimate SNP-based global and local kinship coefficients
on dense genomewide data. It can also use the SNP-based global estimates
to cluster individuals and thus construct pedigree groupings without
pre-existing relationship information.

The input data formats are now more flexible. For example, PLINK binary
data sets (.fam, .bin, and .bed files) are now accepted (see the Mendel
Documentation for details). The SNP definition file can now include
allele names. Analogous to the SNP subset file, there is now a sample
subset file to easily change which subset of individuals should be
included in an analysis. Also, pedigree files without the twin field or
without any pedigree fields can be read directly. These options are set
using the new keyword INPUT_FORMAT.

Many smaller improvements were also implemented. For example, missing
SNP genotypes are now replaced with random ``average'' genotypes when
missing data is not allowed in an analysis. Each selection of a random
genotype is based on the allele frequencies within the data set. The
GWAS output has been expanded and an issue with the Q-Q plot data is
fixed. Also in the GWAS analysis option, one can now set specific values
for the LASSO tuning constants using the command TUNING_CONSTANT, if one
wishes to override Mendel's automatic adjustments. 

In version 13.2 the analysis to rank markers by their informativeness
to distinguish subpopulations, has been extended to allow non-uniform
population contributions. There are also a few minor bug fixes and
other improvements.




 New Features in Mendel Version 12

In this update we have added one new option, Trait Simulation,
and improved several of the existing options. The new option uses either
existing genotype data or the output of the genotype simulation option
to determine trait values. The traits can be either univariate
(simulated by generalized linear models) or multivariate
(simulated by variance component models). Mean effects due to the
genotypes at a major locus can be included in the model.
The list of supported distributional families for the mean effects model
is substantial (see Tables 14.1 and 14.2 in the Documentation).

Genotype simulation now allows various output styles. In addition,
one has considerable control over the degree of missing data
for both trait and genotype simulation. This flexibility is accomplished
with the introduction of four new keywords:: GENE_DROP_OUTPUT,
KEEP_FOUNDER_GENOTYPES, MISSING_AT_RANDOM, and MISSING_DATA_PATTERN.

Since a grand mean (intercept) is required for each trait
in all of Mendel's regression analyses, it is now added automatically.
Thus it is no longer necessary to add commands such as
PREDICTOR = GRAND :: trait
to your model. It does no harm if you continue to include such commands
explicitly.

An important change has been made to the meaning of the weight
that can be assigned to each SNP in the SNP definition file.
These predictor weights are used when building the best regression model
that includes the user-specified number of predictors. A weight
can now be any positive real number. Also, the greater the weight,
the more likely the predictor will be included in the model.
If the keyword UNIFORM_WEIGHTS has the value True, which is the default,
then all predictors with unassigned weights receive weight 1.0.
Otherwise, SNPs with unassigned weights receive weight 1/sqrt{4q(1-q)},
where q is the minor allele frequency at the SNP.
(If you previously included weights in your SNP definition files,
we suggest you now use the inverse of the old weights.)

Other minor changes include the maximum number of loci that can be
combined into a superlocus has been increased from four to five.
The keyword SEED can now be set to the special value TIME,
which sets the seed based on the current time. The value set is
reported in the standard output file.




 New Features in Mendel Version 11

This revision incorporates two new options and several improvements
to existing options. The new options are Maternal-Fetal Genotype
Incompatibility Testing and Inbred Strains Analysis. The first of the
new options facilitates modeling of interactions between maternal and
child genotypes. The classic example is Rh maternal-fetal incompatibility.
The second of the new options represents a new approach to QTL mapping
with inbred strains of mice and other animals. The option implements
a mixed effects model that correctly captures polygenic background,
handles multivariate traits, and copes with pedigrees of arbitrary complexity.
The QTL effect is modeled at the mean level as a vector of regression
coefficients on the strain origin pair (maternal-paternal) imputed
for each animal at the current QTL location along the genome.

To better deal with rare SNPs, the SNP Association option has been
significantly improved. In an expanded version of the SNP Definition file,
for each SNP one can now assign a weight and membership in a group
of SNPs, e.g., a gene or molecular pathway. Default weights can be
based on allele frequencies. The lasso regression can now use group penalties,
which make it easier to understand association at the gene level rather
than the SNP level. The new keyword LASSO_PROPORTION specifies the relative
importance the model puts on individual predictors versus predictor groups.
One may also specify individual predictors, or groups of predictors, to always
retain in the LASSO model by using the keywords RETAINED_PREDICTOR and
RETAINED_GROUP. Finally, in interaction testing, it is now possible
to test for all pairwise interactions in a large base set of predictors.

Among the modification of existing features, we have introduced a better
facility for imposing linear constraints on parameters. See the discussion
of the new keyword PARAMETER_EQUATION in the documentation. The deprecated,
previous keywords PARAMETER_FIXED_VALUE and PARAMETER_EQUATE will continue
to work for now.

The SNP Imputation option now performs both pedigree based and linkage
disequilibrium based genotype and haplotype inference. The combination
of these two approaches leads to more accurate imputation results.
Better imputation in turn leads to better handling of missing data
and presumably more sensitive SNP association tests.





 New Features in Mendel Version 10.0

In this revision several analysis options have been expanded to add
useful new features. For example, both SNP imputation and association
testing now work on X-linked SNPs. The only caveat is that for SNP imputation,
either all or none of the SNPs in each data set should be X-linked.
SNP association testing of case-control data is now much faster, often more
than five times faster. This is accomplished at the expense of calculating
regression coefficients for only at least marginally significant SNPs.
The cutoff p-value significance level can be set via the new keyword
ESTIMATION_THRESHOLD, which has default value 0.001. Also, SNP association
analysis can be performed under any of three dominance models: codominant
(the default), first allele dominant, or first allele recessive.
Alternatively, each SNP can automatically be analyzed under all three models,
and the most significant result reported. The model is chosen by setting
the new keyword SNP_DOMINANCE_MODEL to one of: Codominant, Dominant,
Recessive, or Maximum.

All high-density SNP based analyses now permit filtering the individuals and
SNPs included in the analysis. Filtering is based on genotyping success rates.
The minimum genotyping rates allowed for included individuals and
SNPs are set via the new keywords MIN_SUCCESS_RATE_PER_INDIVIDUAL and
MIN_SUCCESS_RATE_PER_SNP. Both keywords have default values 0.98.
SNPs are filtered first and then individuals, based on the non-filtered SNPs.

The NPL option has been expanded to allow the analysis of quantitative
variables. The new Q-NPL statistic is discussed in the documentation.
For the Location Score analysis option, the trait locus should now be named
using the keyword AFFECTED_LOCUS_OR_FACTOR, and no longer need be included
in the Map file. (Old files with the trait locus indicated by the first locus
in the Map file will continue to work.) The positions analyzed outside
the left and rightmost markers are now set using the new keywords
FLANKING_DISTANCE and FLANKING_POINTS. In a change from past versions,
the values of the keywords FLANKING_DISTANCE and GRID_INCREMENT
now always have the same units as the distances in the Map file,
which recall is set using the keyword MAP_DISTANCE_UNITS. Finally,
for the location score option, a new example data set is added
which illustrates the use of the simple PENETRANCE keyword
to define a penetrance function.




 New Features in Mendel Version 9.0

Significant new analysis options have been added to Mendel to enable
analysis of high-density, genome-wide SNP data. First, new data file
formats are now used by Mendel to efficiently handle the possibly billions
of genotypes in these data sets. The new file formats are discussed in
Section 0.6 of the documentation. Standard, SNP-major PLINK binary pedigree
files, so-called "bed" files, can now be used by Mendel. In addition, we
allow for ordered SNP genotypes through the use of a coordinated SNP phase
file. Conversions between comma-separated Mendel files and these new binary
data files, in either direction, are performed by the new FILE_CONVERSION
analysis option, which is discussed in Section 25 of the documentation.

Two major new analysis options require these binary SNP data files as their
input, SNP_IMPUTATION and SNP_ASSOCIATION. SNP imputation, as described in
Section 23, fills in all missing SNP genotypes based predominantly on linkage
disequilibrium (LD) considerations. This option can also impute all SNP
haplotypes if requested.

SNP association, as discussed in Section 24, tests for association
between SNPs and a quantitative trait or a case-control disease status.
This option performs both univariate SNP-by-SNP analysis and lasso penalized
multivariate analysis. It will test for interactions among the most potent
predictors. You can select this set of predictors or let Mendel do it for you.
If you request higher-order interactions, please keep in mind that Mendel
can be overwhelmed by the sheer number of interactions sought.

Several additions have been made to the TRIM_PEDIGREES analysis option.
Input data for this option can now come from binary SNP data files. In this
case, the results are repackaged into new binary data files. Four new
trimming models have also been introduced, for a total of nine. Each of the
new models extracts specified individuals and puts them in new files as
unrelateds.

Hardy-Weinberg equilibrium testing can now be performed on pedigrees as
part of allele frequency estimation; see the documentation for more
details. There are also two new quantitative transformation functions,
ABOVE_THRESHOLD and BELOW_THRESHOLD. In the transformed variable,
ABOVE_THRESHOLD replaces all values at or above a user-specified threshold
with 1, and all other non-missing values with 0. Use the new keyword
INDICATOR_THRESHOLD to set the threshold value. BELOW_THRESHOLD does the same,
but reverses the values receiving 0 or 1.

For genotypes not listed in the penetrance file, or under control of
penetrance commands in the control file, the default penetrance value is now 1.
As always this can be changed using the keyword DEFAULT_PENETRANCE in the
control file. Regardless of its default penetrance value, Mendel assigns
a penetrance of 1 to each legal genotype of someone missing both a penetrance
in the penetrance file and a phenotype at the locus.

Finally, the map file is no longer necessary if all loci listed in the
definition file should be included in the analysis, and should be treated
as unlinked loci. The sole exception at this time is Location Score
analysis, where the map file is still required since the first locus listed
in the map file is mapped against all the other model loci.




 New Features in Mendel Version 8.0

Version 8.0 implements some major changes that users should bear carefully
in mind. Most importantly, all input file format are by default now
list-directed; thus, data values must be separated by commas, blanks, or
tabs. To restore the previous default of column formatted files, simply set
the keyword DEFAULT_LIST_READ to false in the control file. In
list-directed pedigree files, recall that each person record begins with a
pedigree name. In column formatted pedigree files, each pedigree is
preceded by a pedigree record giving the number of people in the pedigree,
the pedigree name, and optionally the number of copies of the pedigree. 
These defaults can be overridden.

Because later versions of Mendel will be geared to large SNP data sets, we
no longer support the previously named SNP_FILE. This file was used to
identify loci to be combined into super-loci. The corresponding Mendel
option, Combining_SNPs, has been renamed Combining_Loci. In this option,
you now designate the loci to be combined at the bottom of the definition
file. The relevant syntax is similar to the syntax used to describe factors
and variables in the definition file; see Section 0.5.5.4 in the
documentation for details. As described in Section 18, it is also possible
to designate successive super-loci by sliding a window of fixed width
across the map file.

Mendel now allows the map file to list each locus position rather than the
genetic distances or recombination fractions between adjacent loci. When
you use physical distance units of bp, Kb, or Mb, Mendel expects this new
position style. See Section 0.5.6 for more details and example map files.

Mendel now implements Cox's proportional hazards model in penetrance
estimation. See Section 14.4 for more details. In gene dropping, Mendel
will now randomly generate pedigrees with founder sources rather than
allele names labeled in the pedigree output file.  Simply set the new
keyword GENE_DROP_LABELS to true in the control file.  See Section 17.4 for
an example.

The remaining new features of Version 8.0 involve minor changes to the
input files or their processing. For example, using keywords
MISSING_QUANTITATIVE_VALUE and MISSING_VALUE instructs Mendel to interpret
the assigned string as a missing value symbol for quantitative and
non-quantitative fields, respectively. The string assigned to either
keyword can be up to four characters long. When the pedigree file is in
LINKAGE format, the default for MISSING_VALUE is "0", the zero character;
otherwise, the default is blank. The default value for
MISSING_QUANTITATIVE_VALUE is always "-".

To avoid confusion, Mendel now treats the two slash characters, / and \,
interchangeably. In particular, either slash can now be used to separate
alleles in an unordered genotype in any input files. If a value is assigned
to the ALLELE_SEPARATOR keyword, then only that value can be used as the
unordered separator. Also to avoid confusion, the ORDERED_ALLELE_SEPARATOR
keyword can no longer be assigned either slash; its default remains the
vertical bar, |. The keyword AFFECTED now also has two values by default,
"AFFECTED" and "2", so either value can be used to indicate affected
individuals. Again, if a value is assigned to the AFFECTED keyword in the
control file, then only that value will label affecteds. LINKAGE formatted
pedigree files must continue using a "2" to indicate affecteds. Recall that
setting the AFFECTED keyword to "everyone" (case insensitive) will
designate everyone as affected, regardless of the recorded values of the
AFFECTED_LOCUS_OR_FACTOR.

In the chromosome descriptor field for loci listed in the definition file,
Mendel now recognizes "XY" and "YX" as designating an autosomal locus.
Mitochondrial and Y-linked loci are also now recognized, but these loci are
ignored during analysis. More detail on the changes in Mendel's treatment
of this field can be found in Table 0.3 in the documentation. Finally, the
new keyword PEDIGREE_MAX_LINE_LEN takes an integer value that determines
the maximum line length in comma-separated pedigree files; the default
value is 262,144 characters. Setting this value much higher slows Mendel's
initialization phase.

A couple of changes have been made in the minor update to version 8.0.1.
A new pedigree file can be created under model 1 of the mistyping option.
In this new file, in each pedigree that a mendelian error is found at a locus,
all phenotypes at that locus will be left blank in that pedigree. Pedigree
pretrimming is no longer carried out under model 1 of the mistyping option.
Finally, a bug was fixed affecting reading of blank or commented-out lines
in list directed pedigree files. (Thanks to Dr Duffy for catching that bug.)




 New Features in Mendel Version 7.0

In version 7.0 we begin a migration of Mendel toward comma-separated
input files. Setting the new keyword DEFAULT_LIST_READ equal to true
in the control file designates all input files as comma-separated.
This blanket declaration can be overridden for a specific input file
by the usual file-specific command. Eventually all input files will
be comma-separated by default rather than column-specific by default.
Although column-specific files are easier to read, they are painful
to create. Neither choice is ideal, and users should beware of some
of the pitfalls with comma-separated files. For example, Fortran treats
commas, tabs, spaces, and forward slashes (/) as field separators.
This explains the impossibility of retaining the forward slash as the
default allele separator in comma-separated files. The backward slash (\)
is now the default allele separator in this setting rather than our
earlier favored dash.  Because Excel uses dashes and colons for dates
and times, many users found the dash to be a poor choice for an allele
separator. Please keep in mind that, except for file names, neither
internal spaces, internal tabs, commas, forward slashs, backward slashs,
exclamation points, vertical bars, ampersands, nor plus signs should
appear as part of a name in an input file. 

We are also planning to replace the locus and variable files by a
single definition file that contains both types of information.
Although Version 7.0 is backward compatible, we recommend adopting
the definition file in all of your future analyses. For comma-separated
pedigree files, the new default omits pedigree records. Thus, each
person record should begin with a pedigree name. This convention can be
overridden in the control file. The expiration date of version 7.0 and
future versions will no longer be the end of the calendar year. We have
instituted a grace period of three months to allow enough time for
installation of new versions. Proband pedigrees are now allowed in all
likelihood-based options. Finally, it is now possible to bypass the
penetrance file and enter simple, uniform penetrance models from the
control file using the new PENETRANCE keyword. See the end of Section
0.5.11 in the Documentation for detailed instructions.




 New Features in Mendel Version 6.5

The format of the penetrance file has been completely revamped. Our new
penetrance files are easier to understand and more flexible. Although old
penetrance files are now obsolete, we show how to construct acceptable
substitutes in analysis option 14. The new keyword DEFAULT_PENETRANCE
is helpful in reducing the size of a penetrance file. Our more flexible
penetrance format allows us to handle an interesting genetic counseling
problem in option 7 involving the age of onset of a form of hereditary
deafness.

Version 6.5 contains numerous other minor adjustments. Most of these are
internal to Mendel and represent bug fixes and speedups. Other changes are
visible to the user. For instance, it is now possible to transform a trait
dichotomy such as affection status into a 0/1 quantitative variable. This
transformation is necessary in analysis option 14, which operates on
quantitative traits. In analysis option 18 on combining SNPs, the keyword
EQUILIBRIUM_FREQUENCIES now determines whether linkage equilibrium haplotype
frequencies are deposited in the new locus file. Options 11 and 12 now
generate summary files. The locus file is now slightly easier. When all
genotypes are compatible with a phenotype, they do not need to be enumerated
in the locus file. Distances in the map file can now be physical distances,
genetic distances, or recombination fractions, see the documentation for
a list of the options and a discussion of how Mendel transforms between them.
Finally, the keyword AFFECTED can now be assigned the special value Everyone,
which will designate everyone as affected, regardless of the value for
AFFECTED_LOCUS_OR_FACTOR.




 New Features in Mendel Version 6.0

The previous analysis option 15 on Gaussian penetrances has been replaced by
a new option 15 on ethnic admixture. The functionality of the old option 15
is now included as part of option 14 on penetrance estimation. Option 14
itself has been upgraded to accommodate generalized linear penetrance models,
trait censoring, and better ascertainment correction. See the documentation
for a description of how Mendel handles probands and ascertainment correction.
To promote computational efficiency, a limited form of pedigree trimming is
now done automatically as a preprocessing step in most analysis options.
Mendel's default trimming is sketched in the documentation; more extensive
trimming is possible using option 21, including splitting pedigrees into their
constituent nuclear families. In maximum likelihood estimation, it is now
possible to fix, initialize, and bound parameter values in the control file.
The documentation discusses the new keywords implementing these commands.

Analysis option 8 on the gamete competition model and analysis option 13 on
the TDT now permit users to assess transmission distortion by the gender of
the transmitting parent. In option 8 one can also treat transmission to
normal children as complementary to transmission to affected children.
These vague distinctions are made clear in the revised documentation of
the two options. As alluded to above, option 15 facilitates estimation of
ethnic admixture. In contrast to competing software, this is done on whole
pedigree data. The ethnic fractions computed for each individual can be
output to a new pedigree file for use as trait predictors in other options.
Options 19 and 20 now allow for X-linked trait contributions.
Finally, a new command line interface, described in the documentation,
makes it possible to invoke a different control file at run time than the
default file Control.in.




 New Features in Mendel Version 5.7

The new keyword DEVIANCE_VARIABLE has been added to allow ordered subset
analysis with user-specified pedigree ordering; its use and function is
explained in the documentation. Analysis option 6 on allele frequency
estimation has been extensively reworked. In model 2 the most likely genotype
is now given for each ambiguous phenotype. This effectively solves the
haplotyping problem for a random sample of unrelated people. Consider applying
model 2 of option 6 to the SNP combination markers output by option 18.
Depending on the number of populations input, option 6 now evaluates either
Wright's F_ST statistic for population substructure or Watterson's
homozygosity test for selection pressure. Option 11 on genetic equilibrium and
option 12 on cases and controls have been revamped to accept two haplotypes
per person. Each analysis is clearly labeled by data type - haplotype data,
multilocus genotype data, or multilocus phenotype data. The data type is
automatically chosen by Mendel. The current version of Mendel no longer
supports the old method of inputting haplotypes, which asked the user to
create one completely homozygous individual for each observed haplotype.
Option 11 tallies the number of different haplotypes seen in the data and
provides the p-value of this statistic.

A new analysis option, number 22, has been added to provide a test for
association given linkage. This option implements, with slight modifications,
Xiong and Jin's model [71] capturing association and linkage simultaneously
in pedigree data. In particular, it estimates both recombination and linkage
disequilibrium parameters and conducts a likelihood ratio test for association
given linkage.

Several modifications have been made to the use of comma-separated input
files. For example, trailing commas after the last non-blank value are no
longer required, and locus and factor names can now be up to 16 characters
long. Other changes are described in the documentation.

Finally, there were assorted bug fixes. Several of these bugs were reported by
Sue Service whom we thank!


 New Features in Mendel Version 5.5.2

The changes from version 5.5 to version 5.5.2 included bug fixes for issues
reported by Daniel Weeks and Anne-Louise Leutenegger. Thank you for your help!
These fixes do not change any results previously obtained from version 5.5,
they only increase analysis speed and allow Mendel to work on a wider variety
of data sets, especially larger data sets. Also new in version 5.5.2 is a
permutation-based p-value estimation technique for the quantitative trait
association test in option 20.


 New Features in Mendel Version 5.5

The Lander-Green-Kruglyak algorithm for likelihood evaluation has been
completely rewritten to incorporate the innovations described in the
articles [1,43]. As a consequence, we have added the keyword GENE_FLOW_CUTOFF,
which directs Mendel to delete descent graphs of low probability. Option 1
on mapping markers now has two additional models that automate the process of
ordering a large number of markers. Instead of visiting all possible marker
orders, Mendel estimates the recombination fraction separating each pair of
markers and then uses these pairwise estimates to generate a list of promising
candidate orders. The number of candidate orders retained is dictated by the
keyword CANDIDATE_ORDERS. We have added to option 2 an example illustrating
the application of reduced penetrance in mapping a trait locus. Option 3 has
the capacity to output pedigree files displaying inferred genotypes. When it
is analyzing only small pedigrees, Option 19 no longer requires precomputation
of conditional kinship coefficients by SimWalk2. Finally, Mendel now has a new
option, 21, for trimming irrelevant people from a pedigree and splitting a
pedigree into its connected components. These actions can dramatically
accelerate likelihood computation.

[1] Abecasis GR, Cherny SS, Cookson WO, Cardon LR (2001)
Merlin - rapid analysis of dense genetic maps using sparse gene flow trees.
Nature Genet 30:97-101

[43] Markianos K, Daly MJ, Kruglyak L (2001)
Efficient multipoint linkage analysis through reduction in inheritance space.
Amer J Hum Genet 68:963-977

[71] Xiong M, Jin L (2000) Combined linkage and linkage disequilibrium mapping
for genome screens. Genet Epidemiology 19:211-234