buildfw

`buildfw`

Tries to heuristically build a good framework map, i.e., a map such that all alternatives orders have a log-likelihood not within a given threshold of the framework map. The command may also be used to build comprehensive maps (including all markers) by adjusting thresholds.

Synopsis :

The CarthaGene buildfw command is invoked as either one of :

buildfw Options
buildfw keepThres AddThres MrkList MrkTest

Description :

The procedure is an incremental insertion procedure. It starts either by selecting a triplet of loci or by using the list of loci MrkList. The difference in log-likelihood of the two best maps you can build with three given loci is used as criterion. The greater the difference the better the triplet.

Then, each available locus is tentatively inserted at each possible position. If there is no locus that provides a difference in log-likelihood larger than the AddThres threshold, the algorithm stops. If there are many orders for which the difference of log-likelihood is larger than the KeepThres Threshold, they are kept for the next step. The MrkTest flag indicates if post-processing is applied. If it is set to 0 no post-processing occurs. If it is set to 1, each remaining locus is tentatively inserted at each position in the previously built framework map and the difference in log-likelihood between each possible insertion and the best one (denoted by a '+') is reported. If it set to 2, no framework is built in the previous stage, instead the MrkList list of loci is used and the same post-processing as above applies. If the MrkTest equals 0 or 1, the framework's markers become the list of currently active markers.

When MrkTest flag is different from 0, further information on the remaining markers that have not been integrated in the map is provided in a table. Each remaining marker is represented on a line. The information is displayed in 6 group of columns.

BuildFW, remaining loci test :

      |                  |
      |                  |     Lod2pt         Dist2pt
      | 1 2 3 4 5 6 7 8  |  Left<-M->Right Left<-M->Right | 0->N    N->M | Weight Nb:W<AM  Id  | Name
    --|------------------|--------------------------------|--------------|---------------------|-------
M1702 |              + 0 |   2.16   23.97   115.1   2.5   |  680.8 106.7 |     0     2     9   | M1702

From left to right, we get:

the first group of columns displays the name of the marker that has been tentatively inserted (M1702 above)
the second group of columns displays the result of inserting this marker in the framework map. The best possible insertion position is reported by a ``+''. When alternative insertion positions exists with a small difference in log-likelihood (less than KeepThres), they are also reported by the nearest integer that represents this difference (eg ``0'' above corresponds to a difference in log-likelihood between in ).
the third group of columns indicates the relations of the inserted markers with its two neighbors in the best insertion position. The two-points LOD and the two-points distances (Haldane or centiRay depending on the dataset type) with the two flanking markers are displayed (when available).
When merged datasets are used, the two-points distances reported are computed only on informative data-sets and averaged among the informative data-sets. When RH data is used, and because different irradiation level (or mixed RH/genetic data) may have been used, distances are not averaged: only the first dataset merged is used.
the fourth group of columns aims at positioning the remaining markers wrt. the framework map. It first (0->N) indicates the multi-point distance from the left origin of the map to the left flanking marker estimated on the original framework map (before marker insertion). If the marker best insertion is at the extreme left of the map, a - is displayed. Then (N->M), an estimation of the distance from this left flanking marker to the marker inserted is reported. This distance is simply computed by rescaling the multi-point distance between the two flanking markers using the ratio between the two-points distances from the inserted marker to the left and right flanking markers reported in the previous group of columns. If either the left or right flanking markers are undefined, the two-point distance alone is reported. If there is no available information on either of the flanking markers, NA> or NA< is printed (indicating the marker that generates the indetermination).
the $5^{th}$ group of columns successively reports:
- Weight: the sum of the difference of log-likelihood with the best map for all maps within KeepThres of the best map. A large weight indicates a marker whose position is not well defined.
- Nb:W<AM: the number of insertion position where the difference in log-likelihood exceeds KeepThres.
- Id: the marker numerical Id.
the last column gives the name of the inserted marker again (same as first column).

Arguments :

Options : -u to obtain the synopsis of the normal use, -h to print a one line description, -H to print a short help.
KeepThres: the minimum difference in log-likelihood between the best insertion point and the second best insertion point required for the map to be considered in the future.
AddThres: the minimum difference in log-likelihood between the best insertion point and the second best insertion point required for a locus to be insertable. This threshold is also used to filter out the differences in log-likelihood reported by the postprocessing option (only differences lower than the threshold will be reported).
MrkList: an order of markers to start from (may be empty or caontain at least 3 markers).
MrkTest: a flag to select either framework mapping only (0), framework mapping followed by a postprocessing (1) or just the post-processing applied to (2)MrkList.

Returns :

nothing but the set of active markers is automatically set to the framework map built if the MrkTest equal 0 or 1, not 2.

Example :

# we first load a data set
CG> dsload Data/rh.cg
{1 haploid RH 53 118 /homes/thomas/carthagene/test/Data/rh.cg}
# we perform linkage group detection (output omitted)
CG> group 0.3 3
...
# we select the group  10
CG> mrkselset [groupget 10]

# we build a framework map for this group
CG> buildfw 3 3 {} 1

BuildFW, Adding  Threshold = 3.00, Saving Threshold = 3.00.

>>> Delta = 6.89 :

Map  0 : log10-likelihood =   -66.75
-------:
 Set : Marker List ...
   1 : G7 G12 G16

>>> Delta = 6.30 , Id = 32,  Locus = G18 :

Map  0 : log10-likelihood =   -80.72
-------:
 Set : Marker List ...
   1 : G7 G12 G16 G18

>>> Delta = 7.04 , Id = 28,  Locus = G14 :

Map  0 : log10-likelihood =   -93.15
-------:
 Set : Marker List ...
   1 : G7 G12 G16 G18 G14

BuildFW, remaining loci test :
       |            |
       | 2 2 3 3 2  |     Lod2pt         Dist2pt
       | 1 6 0 2 8  |  Left<-M->Right Left<-M->Right | 0->N    N->M | Weigh...
     --|------------|--------------------------------|--------------|------...
    G5 |      3 1 + |  13.70     -      29.6    -    |   91.9  29.6 |     4...
   G17 |      2 + 0 |  18.05   15.66    21.1  25.3   |   72.0   9.0 |     2...
   G13 |    + 3     |  20.41   22.03    14.6  10.6   |   30.4  12.7 |     3...
    G6 |    + 0     |  16.23   20.68    25.7  14.6   |   30.4  14.0 |     0...

# and check its reliability (any map at less than 3 LOD ?)
CG> flips 4 3 0

Single Flip(window size : 4, threshold : 3.00).


Map -1 : log10-likelihood =   -93.15
-------:
 Set : Marker List ...
   1 : G7 G12 G16 G18 G14


 2 2 3 3 2  log10
 1 6 0 2 8    -93.15


# no.