Table 1.
Summary of the simulated data sets.
Data set: The name of the data set; Markers/chr: Number of markers simulated on each chromosome; Total markers: The total number of markers; Genetic map: Number of markers in the genetic map after filtering for informative markers (and the corresponding percentage of all simulated markers); Markers/LG: The average number of markers on each LG (and the marker density range).
Fig 1.
An overview of the BatchMap pipeline.
A) Input data file is read into BatchMap using the function read.outcross2(). B) A two-point table of recombination fractions is created and likelihoods calculated per phase using the function rf.2pts(). C) Markers are assigned to linkage groups (LGs) using the function group(), sequence objects created using make.seq() and LGs are split into parental maps using marker segregation type. D) Marker order is determined using RECORD, with the number of iterations and threads set using the function record.parallel(). E-F) The genetic map is calculated in centiMorgans by first using the function pick.batch.sizes() to split the LG into even groups and then using function map.overlapping.batches() with the predefined batch size. All pairwise marker phases are set within each batch and overlapping markers are fixed in phase from the previous batch. When all marker pairs are fixed in phase the original OneMap function map() is called to calculate the distance between markers in cM using either the Kosambi or Haldane mapping function. G-H) An optional ripple function, ripple.ord(), can be set in map.overlapping.batches() to consider all single pairwise swaps between markers in a sliding window of predefined size along the LG. When the best order (based on likelihoods) within each window is set, map calculation will be carried out as in E-F).
Fig 2.
Summary statistics of all evaluation runs for OneMap and BatchMap.
In order to control for random effects during ordering and grouping, those factors were estimated once using RECORD and supplied to all runs. A) Boxplots (N = 6) of linkage map sizes in centiMorgans for all evaluations. B) Hours elapsed per calculation for each pseudo-testcross of three linkage groups between 724 and 794 markers in density. C) Boxplots (N = 6) of linkage map likelihoods for all evaluations. D) Frequency polygons for all evaluations summarizing the distance of each marker to its true position. All versions except BatchMap_ripple are identical and hence overlap. BatchMap_ripple was run with a window size of five, which is indicated by the vertical dashed bars.
Fig 3.
Accuracy of estimated marker order (using 33 iterations of RECORD) for the four data sets.
Distribution of distance from true position summed over 10 LGs each for sim7.5k, sim10k, sim 15k and sim20k.