Map Manager QTX, cross-platform software for genetic mapping
Kenneth F. Manly, Robert H. Cudmore, Jr., Jane M. Meer
Cellular and Molecular Biology, Roswell Park Cancer Institute, Buffalo, New York 14263, USA
Received: 2 July 2001 / Accepted: 20 August 2001
Abstract. Map Manager QTX (QTX) is software for analysis of
genetic mapping experiments in experimental plants and animals.
It includes functions for mapping both Mendelian and quantitative
trait loci. QTX is an enhanced version of Map Manager QT, re-
written with the aid of cross-platform libraries (XVT, Boulder
Software Foundry, Inc.), which allow it to be compiled for mul-
tiple computer platforms. It currently is distributed for Microsoft
Windows and Mac OS and is available at http://mapmgr.
Much of the design for QTX, both analysis and user interface,
derives from Map Manager QT (Manly and Olson 1999). Both
programs offer similar data entry and display for Mendelian mark-
ers and quantitative traits. Both calculate two-point linkage and
map distances for Mendelian markers, and offer single-locus as-
sociation, simple interval mapping, and composite interval map-
ping for quantitative trait loci. This report will focus on differences
between the two programs. These differences, described below,
include additional methods for QTL analysis, improved methods
for ordering marker loci and for defining marker maps, support for
additional cross designs and mapping functions, and methods for
defining and using subsets of data.
QTX is still under development. Current plans for further de-
velopment and opportunity to comment on those plans are avail-
able at http://mapmgr.roswellpark.org/mmQTX.html.
Results and Discussion
QTL analysis methods. The development of QTX has removed
some of the limitations of Map Manager QT. QTX supports
crosses with dominant markers and offers a choice of three map-
ping functions, Haldane (Haldane 1919), Kosambi (Kosambi
1944), and Morgan (Crow 1990).
Estimation of missing marker data. The expected QTL effects
of missing marker genotypes are estimated from the genotypes of
adjacent markers by a Markov-chain method as described by Jiang
and Zeng (1997). For efficiency, a set of conditional probabilities
is calculated for each unknown genotype and stored with the ge-
notype. These conditional probabilities can be combined, accord-
ing to Bayes theorem, with expected Mendelian frequencies for
each genotype to estimate a QTL effect either at the unknown
marker or in the intervals flanking the unknown marker. For in-
terval mapping, these conditional probabilities are calculated once
and used repeatedly to estimate QTL effects at various positions in
the intervals flanking an unknown.
This method extends to composite interval mapping. To imple-
ment composite interval mapping, Map Manager provides for each
trait a list of loci and other covariates called the “background” list.
The user chooses loci to add to this list according to their signifi-
cance in single-locus association tests. Loci added to this list are
used in multiple regression to (partially) represent the effect of a
nearby quantitative trait locus while mapping another locus else-
where in the genome. Marker loci in this background list are
references to the original loci, which remain at the proper location
in a chromosome where adjacent loci are available to support
estimation of missing marker genotypes.
Traits as covariates. It is often useful to be able to correct
traits for the effect of other continuous-valued traits, such as age or
weight. To accomplish this, QTX allows traits to be included in a
background list as well as marker loci. Traits in the background list
are included as covariates in a multiple regression model. This
allows correcting the trait of interest for any continuous indepen-
Weighted least-squares regression. When recombinant inbred
strains are used for QTL mapping, multiple individuals are avail-
able for each genotype, and these can be used to determine a trait
variance for each line. Trait variances for strains may differ sig-
nificantly, violating the prerequisite for least-squares regression.
Weighted regression allows these variances to be used, as de-
scribed in the Appendix. To use this feature, the user calculates
means and variances for a phenotype and enters each set of values
as a trait in QTX. In the dialog that initiates QTL mapping for the
trait that represents phenotype means, QTX allows a second trait to
be specified as having phenotype variances.
Marker analysis methods
Automatic construction of marker maps. A common problem
at the beginning of a mapping project is to create ordered linkage
groups from a group of unassigned, unordered loci. A new menu
item called Make Linkage Groups performs this task. It starts a
new chromosome with the most tightly linked pair of loci available
and then uses Distribute (described below) to add as many other
loci as possible. It repeats this procedure until no further pair of
linked loci is available among the remaining loci. To allow for the
possibility of handling large groups of new loci, Make Linkage
Groups repeats the above procedure up to three times, at decreas-
ing stringency levels. The first repetition uses a very stringent
criterion for linkage, which will transfer a significant number of
markers only in large data sets. This method is derived from the
rapid chain delineation method described by Doerge (1996).
The method used by Make Linkage Groups is also available as
a separate menu item called Distribute. This method is designed to
move new loci from a chromosome (the source chromosome) into
their best position in a group of chromosomes (the destination
chromosomes) in which the loci are already ordered. The best
Correspondence to: K.F. Manly; E-mail: email@example.com
Mammalian Genome 12, 930–932 (2001).
© Springer-Verlag New York Inc. 2001