szpiech
commented
1 year ago Hello,
Thanks for your kind words. Let me try to answer below.
For the vcf files we use, is it enough to remove variants with minor allele count less than 2 (—mac 2) and filter out anything but biallelic?
In my testing I've found that not filtering any sites for MAF works best for the XP statistics, but yes you should filter out anything that isn't biallelic.
Also I wonder how the sample size affects the XPEHH scores. If there is a large sample size difference between the reference population and the focal population, is this a problem?
I haven't carefully explored this question, but my hunch is that if you have a sample that is much smaller than the other, the raw XP statistic would be slightly biased in favor of that population. However, this should be a genome-wide mean effect, so normalizing ought to at least take care of that shift. However, I'm not sure how much it might influence power.
The other question is: What does the large values for the scores mean? Do they reflect the strength of the selection signature since a larger genomic region is influenced by the sweep?
An extreme negative score means that there are relatively long high-frequency haplotypes in that region of the genome in your reference population compared to your other population. Extreme positive scores mean the same thing but in the other population compared to the reference. While I think selection coefficient is likely related to the score, the precise relationship is not entirely clear. I would hesitate to draw too strong a conclusion based on the magnitude of the scores alone.
My last question: For the search of clusters with extreme scores implemented in norm, as far as I understood you state "approx percentile for gt threshold wins" column represents approximate percentile of scores greater than 2 in that window. While plotting these extreme score regions in genome, after identifying these windows with extreme scores with norm nonoverlapping command, is it correct to get the normxpehh values of these windows (from the norm result without window) and plot them as a function of position?
Sure, you can do this. See figure 6 in https://academic.oup.com/evlett/article/5/4/408/6697682 for an example.
Hope this helps,
Zachary
On Thu, Sep 21, 2023 at 3:04 AM cgdmkns @.***> wrote:
Dear @szpiech https://github.com/szpiech,
Thank you for developing this software. I'm new in this and trying to use XPEHH for my sister taxa, on which I have some questions. For the vcf files we use, is it enough to remove variants with minor allele count less than 2 (—mac 2) and filter out anything but biallelic? Also I wonder how the sample size affects the XPEHH scores. If there is a large sample size difference between the reference population and the focal population, is this a problem? The other question is: What does the large values for the scores mean? Do they reflect the strength of the selection signature since a larger genomic region is influenced by the sweep? My last question: For the search of clusters with extreme scores implemented in norm, as far as I understood you state "approx percentile for gt threshold wins" column represents approximate percentile of scores greater than 2 in that window. While plotting these extreme score regions in genome, after identifying these windows with extreme scores with norm nonoverlapping command, is it correct to get the normxpehh values of these windows (from the norm result without window) and plot them as a function of position? Sorry if this is something silly to ask : )
Thank you! Cigdem
— Reply to this email directly, view it on GitHub https://github.com/szpiech/selscan/issues/104, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABAKRQTBP6XUDDN3DBQ2PD3X3PRHRANCNFSM6AAAAAA5BBGJTU . You are receiving this because you were mentioned.Message ID: @.***>
cgdmkns
Dear @szpiech,
Thank you for developing this software. I'm new in this and trying to use XPEHH for my sister taxa, on which I have some questions. For the vcf files we use, is it enough to remove variants with minor allele count less than 2 (—mac 2) and filter out anything but biallelic? Also I wonder how the sample size affects the XPEHH scores. If there is a large sample size difference between the reference population and the focal population, is this a problem? The other question is: What does the large values for the scores mean? Do they reflect the strength of the selection signature since a larger genomic region is influenced by the sweep? My last question: For the search of clusters with extreme scores implemented in norm, as far as I understood you state "approx percentile for gt threshold wins" column represents approximate percentile of scores greater than 2 in that window. While plotting these extreme score regions in genome, after identifying these windows with extreme scores with norm nonoverlapping command, is it correct to get the normxpehh values of these windows (from the norm result without window) and plot them as a function of position? Sorry if this is something silly to ask : )
Thank you! Cigdem