Question on --max-gap allowance

[m-r-grundler]

Hello and thank you so much for this great software!

I have a question regarding gap allowances for XP-nsl and XP-ehh calculations - the max-gap for my data needs to be set to around 900,000. I'm wondering if this is an acceptable distance to include in calculations or if there is a threshold at which it's better to skip? My apologies if this is detailed somewhere in the paper or documentation but I haven't been able to find it, nor reason it out on my own as I'm quite to new to this.

Many thanks for your time and valuable insight!
Maggie

[szpiech]

Hi Maggie, Is this because 900kb is the largest gap you observe in your data, or do you have very sparse information across your genomes? Generally, I don't recommend setting max-gap to the largest observed gap in your data, as this parameter is specifically intended to prevent possible biases by integrating over large gaps (this could easily inflate your scores in that region). If you are dealing with sparse data and need to raise max-gap in order to get usable amounts of data, you might consider calculating the distribution of gap sizes, and choosing something that is larger than ~90% of your gaps (I chose this number arbitrarily, it would be wise to examine your data and come up with a number yourself). You might want to consider modifying gap-scale as well. Let me know if you have more questions, Zachary

…

On Wed, Jun 18, 2025 at 12:45 PM m-r-grundler ***@***.***> wrote: *m-r-grundler* created an issue (szpiech/selscan#145) <#145> Hello and thank you so much for this great software! I have a question regarding gap allowances for XP-nsl and XP-ehh calculations - the max-gap for my data needs to be set to around 900,000. I'm wondering if this is an acceptable distance to include in calculations or if there is a threshold at which it's better to skip? My apologies if this is detailed somewhere in the paper or documentation but I haven't been able to find it, nor reason it out on my own as I'm quite to new to this. Many thanks for your time and valuable insight! Maggie — Reply to this email directly, view it on GitHub <#145>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABAKRQUNCQWO23M5R7VL4ED3EGJRLAVCNFSM6AAAAAB7TU2XKSVHI2DSMVQWIX3LMV43ASLTON2WKOZTGE2TONJWGQ2TKMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>