First, you must build your map under CarthaGène. Then use the graphical interface of CarthaGène (CGW.tcl). In the "Maps" menu, select the "Graphical" item. You can configure how the map(s) are rendered using the contextual menu (right mouse button). Then print the map to a file as Postscript file.
This file can be either directly included in your publication or edited. Under Windows this can be achieved using a dedicated editor such as Adobe Illustrator. Under Unix, you can translate the Postscript to an editable file format using the "pstoedit" utility. Assuming that the Postscript file is "map.ps", the translation to SVG (Scalable vector Graphics) in "map.svg" can be achived as follows:
> pstoedit -f plot-svg map.ps map.svg
This file can then be edited under a free SVG editor such as "Inkscape" and exported to formats that can be included under most word processors. Both "pstoedit" and "inkscape" are available as packages under most Linux distributions. Ask your system administrator.
Yes. We'll use the dataset rh1.cg provided in the distribution in the "data" directory to give you an example:
CG> dsload Data/rh1.cg
{1 haploid RH 13 118 /homes/thomas/CartaGene/dev/test/Data/rh1.cg}
CG> mrkselset [mrkids {MS5 MS1 MS15 MS4 G37}]
(mrkids will transform the marker names in to numerical
ids for mrkselset that simply set the current marker
selection used)CG> sem Map -1 : log10-likelihood = -121.24 -------: Set : Marker List ... 1 : MS5 MS1 MS15 MS4 G37
flip,
polish or greedy to check the
validity of the map. Have a look to
cgrobustness in the documentation too.
Note that you can also give an initial order to
buildfw for building framework maps (as a forced
starting order). This is all explained in the manual (outbred
section). Remember that phase MUST be known. Imagine the
parents (F and M) are F0|F1 x M0|M1= A | B x A |
C. The sibling can be:
A | A = F0|M0 = code 1 A | C = F0|M1 = code 2 B | A = F1|M0 = code 4 B | C = F1|M1 = code 8Warning. This is phase dependent. If I change the pahe known genotype of the F(ather), we get
F0|F1 x M0|M1= B | A x
A | C and the code are:
A | A = F1|M0 = code 4 A | C = F1|M1 = code 8 B | A = F0|M0 = code 1 B | C = F0|M1 = code 2
If all the markers follow this schema, you are faced with
a classical backcross situation. Use the "backcross" data
type. If this occurs inside an outbred situation (phases known
!) and if I assume the following phases F0|F1 x M0|M1=A
| B x B | B, then
B|B: (F1|M0 or F1|M1) which means code c A|B: (F0|M0 or F0|M1) which means code 3The two codes are inverted if the heterozygous parent phase changes.
No. In the current state of things, you have to infer the phases on your own. A software like CRIMAP with the chrompic command can help. More recently, Dustin Cartwright has produced the phasing program whose output is compatible with Carthagene and which can be used when many offsprings are available.
Yes. CarthaGene contains a complete programming langage called Tcl and you can use all the facilities of this langage (see the TCL site for exhaustive documentation). For executing commands from a file called for example comfile, simply type the following command in CarthaGène window:
source comfile
Well, CarthaGène on its own does not have "chromosome-oriented" command (except for linkage group identification). You can again rely on the programming langage (see the TCL site for exhaustive documentation) that hosts CarthaGène to work on a chromosome by chromosome basis. You can use variables to store the list of markers in each group and then use "mrkselset" to shift from one linkage group to the next chromosome.
CG> dsload mouse.raw
{1 intercross 308 46 /home/tschiex/Dev/carthagene/distrib/data/mouse.raw}
CG> group 0.2 3
...
CG> set chrom1 [groupget 23]
5 12 218 270 293 300 274 232 171...
CG> mrkselset $chrom1
CG> ....
If you don't want to redo this again and again, store just the
corresponding commands in a text file and you will be able to
reexecute it easily (see question above).
CarthaGene integrates several specific versions of the EM algorithm for handling missing data. For intercross, outbreds and diploid radiated hybrids, this is essentially a classical EM algorithm implementation for HMM. For backcross (and related pedigree like RIL) and for haploid radiated hybrids, there is a specific EM algorithm that can run 1 or 2 orders of magnitude faster than traditional EM implementations (without loss of precision). See this paper. When 2 data sets are merged with
dsmergen, all the markers typed in a data set but missing in another one are considered as missing data in this data set. The E (expectation) phase of EM is distributed in each data set (depending on its type) and the sum of expectation is used to estimate the parameters in the M phase. When 2 data-sets are merged using
dsmergor, a traditional EM algorithm is used independently on each dataset and the loglikelihood are cumulated together. For a given markers order, this yields an overall loglikelihood for the merged dataset and estimated parameters for each dataset.
Yes. This is feasible and easy to achieve under Unix/Linux. You have to launch Carthagene with standard output and input assigned respectively to two named fifos. An 'R' script "skeleton" to do this is available under the "Utilities" link on the left menu (Thanks to Matthieu Falque, from INRA Moulon for making this available to the community).
Note that the 1.3 version of CarthaGene is now much faster on startup as it does not recompute 2pts measures (LOD, distances) on startup but use a cache file and multi-threading to handle 2pts measures and grouping.
Doubled Haploids (DH) are equivalent to backcross data. They should be handled exactly as such using the "data type f2 backcross" header type and "H" and "A" to encode the two possible genotypes ("-" being used for unknowns).