We have just released version 2.0 of the LAMARC program (part of the
LAMARC uses genetic data from a population to estimate several population
parameters: size of subpopulations, subpopulation growth rates, migration
rates among subpopulations, and overall recombination rate. The algorithm
is Metropolis-Hastings Markov Chain Monte Carlo; we search through a wide
space of genealogies representing the possible histories of the data.
Suitable data for this program includes DNA or RNA sequences, SNPs,
microsatellites, or elecrophoretic alleles. Phase may be known or
(with some loss of power) unknown.
A serious bug existed in previous versions of Lamarc (not including
Recombine) that could lead to failure to adequately deal with DNA or
SNP markers of unknown phase. Please disregard any earlier results
with DNA or SNP markers of unknown phase and re-run the analysis.
Version 2.0 adds:
Bayesian Analysis. LAMARC can now run a Bayesian analysis instead of
performing a Frequentist one. LAMARC supports both linear and
logarthimic priors with arbitrary boundaries. Users are strongly
encouraged to set the prior boundaries appropriately for their datasets.
Converter Graphical User Interface. A GUI that replaces the current
text menu converter is now available for all supported platforms. The
GUI is a beta-test release which works poorly on X11 windowing
environments. The text-based file converter is still available.
Constrained Parameters. LAMARC can now combine subsets of parameters
for a particular force into groups. Each group can either be held
constant at a given value or constrained to all be equal to each other.
Runs which estimate fewer independent parameters will use their search
time more efficiently, but incorrectly constraining a parameter may
bias the estimate of all other parameters.
Relative Mu and Ne. LAMARC now handles analysis of data with
known differing rates of mutation or population size: for example,
a run can correctly incorporate both mtDNA and nuclear DNA. It does
not currently allow for the specification of a distribution of rates or
Performance improvements. Maximization of the likelihood surface
now succeeds on a wider range of surfaces. A new arranger is available
for use that just lengthens and shrinks all the branches in the tree;
this may increase the effectiveness of the search, particularly in
Interface Cleanup. Models of marker evolution may now be set/changed
in the menu.
LAMARC is written in C++. We distribute source as well as executables for
Linux, Windows and MacOS. To obtain the program or simply learn more
about it, visit our web site:
or our anonymous ftp site:
evolution.genetics.washington.edu, directory /pub/lamarc
The LAMARC development team: lamarc at gs.washington.edu
Department of Genome Sciences
University of Washington
Seattle, WA 98195-7730