A Mapping Software Workshop was held in conjunction with the first
annual Plant Genome Conference in San Diego Nov 8-11 1992. In
response to requests from the community, we are posting in this notice
the abstracts of the presentations from that workshop.
The Conference brought together plant genome researchers from around
the world, with interests in a number of agronomic and model species,
including maize, wheat, soy, rice, pine, and arabidopsis.
The Mapping workshop included demonstrations and discussions
about software for constructing genetic maps, integrating them into
database environments, and presenting graphical pedigrees and other
map-related graphics. More than 150 attended the workshop. Presenters
were: Jeanne Romero-Severson, Piet Stam, Steve Lincoln, Stan Letovsky,
Sam Boutin, Ken Manly, Nat Goodman, Steve Knapp, and Ben Liu.
Organizers were Mary Berlyn of Yale University, Ed Coe of U.Missouri
and Stan Letovsky of Letovsky Associates, New Haven. Informal
discussions during and after the workshop among developers and users
and the workshop organizers made clear the desirability and
willingness of developers to coordinate their efforts and make the
software compatible and easily accessible. Possibilities for
presenting a number of the different programs "under a single shell"
were discussed and proposals for accomplishing this are being explored
Abstracts of Presentations:
Centre for Plant Breeding and Reproduction Research (CPRO-DLO),
PO Box 16, 6700 AA Wageningen, The Netherlands,
Department of Genetics, Wageningen Agricultural University,
Dreyenlaan 2, 6703 AH Wageningen, The Netherlands.
JoinMap constructs and merges genetic linkage maps. Data to be fed are
raw segregation data (i.e. coded genotypes) and/or listed pairwise
estimates of recombination frequencies. Several segregation types (F2,
BC, RIL of any generation, offspring from single pair matings) may
occur in a single input data file. Format of raw data input files
closely resembles MAPMAKER format. Map construction is
non-interactive, but a range of options (critical levels of LOD score,
mapping function, suggested orders, etc.) can be supplied, either from
keybord or from a settings file. JoinMap is especially useful to merge
maps obtained from distinct experiments, and published recombination
frequencies. Available are versions for IBM PC (MSDOS), VAX systems
(VMS), Macintosh and SunSparc Workstations (Unix).
Surveyor, the mapping program used at Agrigenetics, uses a three-tsep
procedure to create a genetic map. The procedure is:
1) the determination of the probability of linkage between any two
2) the estimation of genetic linkage once linkage is suspected
3) the determination of the "best" linear order of a set of genetic
markers once all pairwsise linkages are estimated.
Surveyor employs an analysis of chi-squared distributions to determine the
probability of linkage. Once linkage has been demonstrated, the maximum
likelihood formulas appropriate to the population structure are used to
estimate a genetic linkage. The standard error associated with the estimate
is also calculated.
The determination of the best linear order proceeds by minimizing
the overall standard error of an order within parameters set by the
scientist. These may be either a set number of groups ("chromosomes")
or a maximum linkage distance between markers (1-50 map units). Determination
of order, once the parameters are set, proceeds noninteractively.
CPROP: A Rule-Based Program for Constructing Genetic Maps
Stanley Letovsky and Mary B. Berlyn
Letovsky Associates Department of Biology and School of
286 West Rock Ave. Forestry & Environmental Studies
New Haven, CT 06515 Yale University, New Haven, CT 06520
letovsky at cs.yale.eduberlyn at yalemed.bitnet
CPROP, for Constraint Propagator, is a program that assembles genetic
maps from distance and ordering constraints derived from crosses,
restriction maps, sequencing or other data. CPROP utilizes a
constraint representation of mapping information that supports
explicit representation of knowledge and ignorance about order,
orientation and distance of markers and marker clusters. It automates
rules of inference that geneticists use to derive new facts about order and
distance between loci from the input data. It produces maps that show
not only what is known, but what is not known.
Our first version of CPROP, described in Genomics 12, pp.435-446
(1992), imposes a strict model of additivity of distance, within error
bounds, on genetic data, and reports a contradiction if the data fail
to satisfy this requirement. In the standalone version of the program
these contradictions must be resolved manually; while in a version
integrated with the Coli Genetic Stock Center database, they can be
resolved interactively, via either manual changes or by automated
union or averaging. In the database version the original data and a
record of conflict resolution are preserved in the database.
We have found that in practice, metric conflicts in many kinds
of genetic data occur all too frequently and ad hoc resolutions
accumulate. Our recent work has focussed on loosening the additivity
requirements, using a novel method of integrating a set of metric
constraints on a partially ordered set of loci. The method is based on
a physical analogy of measurements as springs, whose rest lengths and
stiffnesses are determined by the means and standard errors,
respectively, of the corresponding measurements. Intermarker
distances are optimized with respect to the data by allowing this
network of springs to relax until the net force on each locus is zero;
standard errors of the resulting distances are determined by computing
the stiffness of a spring equivalent to the network for each marker
pair. We have a preliminary implementation of this approach, in which
the distance optimization process is embedded as a subroutine within
the larger control structure of CPROP, which also applies rules for
inferring order and distance. The resulting version of CPROP is much
less vulnerable to metric inconsistencies and facilitates integration
of repeated or mutually constraining measurements on a set of loci.
GMENDEL Version 2.0
Ben-Hui Liu * and Steven J. Knapp **
* Department of Forestry
North Carolina State University
Raleigh, NC 27695-8203
** Department of Crop and Soil Science
Oregon State University
Corvallis, Oregon 73331
GMENDEL 2.0 is a program to facilitate genetic mapping. A coding
system is used in the program that enables it to handle a broad range
of data. We have put likelihood functions for commonly used mapping
populations in a default database (dfefreq.data) and likelihood
functions for mapping populations of mating between heterozygous
parents in a complex database (complex.data). Users can create
likelihood functions for their specific problems and put them into the
data bases by following the coding system. A simulated annealing
algorithm (SA) is implemented in the program which makes the locus
ordering fast. The program has a large data capacity and can handle
multiple population mapping problems. Some of the specifications of
the program are:
Computer language: FORTRAN
Computer required: UNIX system with at least 8 MB memory
Maximum number of loci: 1000 (8 MB memory)
Maximum linkage group size: 400 (8 MB memory)
Maximum number of observations: 1000 (8 MB memory)
Default type of mapping population: F2, BC, RI, DH, heterozygous F1
Mapping populations with user specified likelihood:
multiple alleles probelms, segregation distortion, et al.
Number of populations: 1 to 20
Locus ordering algorithm: Sum of adjacent recombination frequency (SAR)
Linkage group size >= 10: SAR with SA
Linkage group size =< 9: SAR with direct evaluation
Multi-point recombination frequency estimation: EM algorithm
ENHANCEMENTS TO MAPMAKER VERSION 3.0
Stephen E. Lincoln, Mark J. Daly, and Eric S. Lander
Massachusetts Institute of Technology, Center for Genome
Research, and Whitehead Institute for Biomedical Research
Nine Cambridge Center, Cambridge, Massachusetts, 02142
We announce the beta release of MAPMAKER/EXP version 3.0 and
MAPMAKER/QTL version 2.0. MAPMAKER/EXP is a full multipoint analysis
computer program for de novo map construction of Mendelian markers
segregating in experimental crosses (please note that MAPMAKER/EXP
does not analyze 3-generation nuclear pedigrees as the original
MAPMAKER did - these features remain available in the previous
version). MAPMAKER/QTL is our package for mapping genes underlying
quantitative traits using a genetic map.
MAPMAKER/EXP now adds the following features.
* An algorithm for detecting typing errors. (see Lincoln and Lander,
Genomics 14: 604-610)
* Many features for automating analysis and storing results,
particularly useful with very large data sets.
* The basic structure of a database interface, customizable by
* Many Bug Fixes and feature improvements.
* Support for Recombinant Inbred and F3-Self populations, in addition
to F2 and BC1. RI analysis is performed a computationally efficient
* The ability to run on PC compatibles running DOS. MAPMAKER requires
a 32-bit 386 or 486 CPU with a math co-processor (e.g. not an "SX"
model) and (depending on data set size) at least 8+ Megabytes of
memory, and 10 Megabytes of free disk space. MAPMAKER is compatible
with, but does not make use of use, MS-Windows 3.1.
MAPMAKER/QTL 2.0 is a bug-fixed improvement to MAPMAKER/QTL designed
to work with MAPMAKER/EXP 3.0. A major release of MAPMAKER/QTL,
version 3.0, is planned which will also add support for RI and F3
populations, and will calculate LOD scores for incompletely penetrant
genes controlling qualitative (+/-) traits.
MAPMAKER/EXP has been extensively tested as part of our lab's
construction of dense genetic maps of the Mouse and Rat, and in
mapping genes underlying various traits in those species (see e.g.
Dietrich et al. Genetics 131:423-427 (1991), Jacob et al. Cell
The beta distribution of these packages for Suns and PC's is being
handled by Mark Daly (617) 258-5135, FAX 258-6505,
mjdaly at athena.mit.edu. Source code is available by FTP from