The genome-wide sliding window estimates of gene flow resulted in only six distinct values.

[kuangzhuoran]

I plotted the distribution of gene flow (me) along the entire chromosome:

[kuangzhuoran]

I ran gimble parse, gimble blocks, gimble windows, gimble tally, gimble makegrid, and gimble gridsearch, and then used gimble query to extract the results of the gridsearch。

[kuangzhuoran]

After gimbleprep, my command is :

1. gimble parse -v in.vcf -b intergenic.bed -g ref.fa -s samples.csv -z Chr5
2. gimble blocks -z Chr5.z
3. gimble windows -w 2000 -s 2000 -z Chr5.z
4. gimble tally -l windows_kmax2 -t windows -z Chr5.z
5. gimble makegrid -z Chr5.z -l Chr5.AB -m IM_AB -b 64 -r A -u 3.03e-9 -A 5000,300000,6,log -B 5000,300000,6,log -C 5000,300000,6,log -T 10000,1180000,6,log -M 1e-7,1e-3,6,log
6. gimble gridsearch -z Chr5.z -g makegrid/Chr5.AB -d tally/windows_kmax2
7. gimble query -z Chr5.z -l gridsearch/windows_kmax2/Chr5.AB --sliced me

I suspect that the issue may have occurred during gimble makegrid, specifically with the -M 1e-7,1e-3,6,log parameter.
The third value of the -M argument was set to 6, which resulted in only 6 distinct values in my output.

Maybe I should have set a higher number?
How should I configure it to achieve an effect like the one shown in the figure below?

[kuangzhuoran]

When I use freeBayes to identify SNPs, each sample is identified separately and then merged together, which actually results in a small number of SNPs. That may be the reason

I have retrieved a new SNP dataset and it is currently running。

[kuangzhuoran]

This only depends on the “-M 1e-7,1e-3,6,log”.
If I change the "6" to "50", I will obtain 50 distinct values. (-M 1e-7,1e-3,50,log)

There are so many windows get the max value.
Is this situation correct?

[GertjanBisschop]

As always @KLohse might be able to provide more biological insight, but from an implementation perspective, it is true that if you increase the number of gridpoints (from 6 to 50), then the likelihoods for a larger number of values for the parameter of interest will be computed. Quite often when your hitting the boundary of your grid, this means that the optimal value is probably bigger than your current max. It might be a good idea to first increase the size of your grid (larger range of values) to first explore what the range of values is you might expect to get.

[kuangzhuoran]

Thank you for your reply, in addition to expanding the upper limit,

1. would it be better to set the me here from “log” to “lin” perhaps?

2. How do I get the range of possible values?
   I was thinking of just setting it to an extremely large value like [1e-7, 1].