help with defining windows - blocks or bp?

[kakapocy]

I am running on gimble v1.0.3

I understand that windows are defined as number of blocks, so -w 500 would give me around 50,000bp (~500*64 callable sites), which is also confirmed by:

 gimble windows --help
usage: gimble windows                    -z <z> [-w <w> -s <s>] [-f] [-h]

    [Options]
        -z, --zarr_file=<z>              Path to existing GimbleStore
        -w, --blocks=<w>                 Number of blocks in windows [default: 500]
        -s, --steps=<s>                  Number of steps (blocks) by which windows are shifted [default: 50]
        -f, --force                      Force overwrite of existing data
        -h, --help                       show this

I ran this:

gimble blocks -z ${i}_analysis.z --block_length 64  --block_span 128
gimble windows -z ${i}_analysis.z -w 500 -s 100
gimble info -z ${i}_analysis.z

Then proceed to gridsearch on windows. But when I ran query and got the .bed file it shows me windows far smaller than my expected window size. Example

# sequence      start   end     window_idx      Ne_A    Ne_A_B  Ne_B    T       me      lnCl
Chr18   0       192     0       190909.0909090909       10000.0 10000.0 522111.0        0.0     -263.54488115419304
Chr18   0       239     1       190909.0909090909       10000.0 10000.0 522111.0        0.0     -281.6188818526653
Chr18   38      260     2       190909.0909090909       10000.0 10000.0 522111.0        0.0     -278.00408171297084
Chr18   64      320     3       190909.0909090909       10000.0 10000.0 522111.0        0.0     -297.1575259187862

Should -w and -s be the actual length in bp instead of number of blocks?

Thank you very much for your help!!

[KLohse]

Please note that -w and -s are specified in terms of number of blocks in heterospecific sample pairs, i.e. you need to account for the combinatorics, when choosing these -w and -s. E.g. assuming you have block =s o 64 bases and 10 samples from each pop/taxon and wish to analyse windows of 500 blocks worth of sequence, i.e. ~ 32kb with a 20% shift between successive overlapping windows; you'd specify -w 50000 -s 10000. This accounts for the 10*10 sample pairs, i.e. 500 * 100... Hope this helps!

[kakapocy]

Thank you very much for the explanation!