08.ehh_stats.md

Selective sweep detection based on extended haplotype homozygosity

We applied extended haplotype homozygosity (EHH) based methods to detect genomic regions under recent positive selection. The tool selscan was used to scan for such regions by calculating three EHH-based statistics, iHS, XP-EHH, and XP-nSL. While iHS is calculated for each population individually the XP (cross population) type statistics both consider pairs of populations.

Since there is no recombination or linkage map available for A. digitifera we used use physical distance instead of genetic distance in all scans. The phased vcf file produced by SHAPEIT was used in all selscan runs, and scans were performed on scaffolds with more than 10 phased SNPs (464 scaffolds). Several subsets of this file were created in order to facilitate scans within and between various combinations of populations. All such subsetting was performed using bcftools version 1.9 as follows.

bcftools view -S {pop} -r {chr} Adigi.v2.indv74_phased.vcf.gz | bgzip > {pop}.{chr}.vcf.gz

Running selscan and norm

We calculated iHS scores for each population separately using selscan 1.3 as follows;

selscan --ihs --vcf {pop}.{chr}.vcf.gz --pmap --threads 10 --out ihs_out/{pop}/{chr}

XP-EHH and XP-nSL scans were conducted in a way that allowed each population to act as a focal population {pop} with the remaining two populations as a combined reference {ref}. Commands used to perform these calculations were as follows;

selscan --xpehh --vcf {pop}.{chr}.vcf.gz \
  --vcf-ref {ref}.{chr}.vcf.gz --pmap \
  --threads 10 --out xpehh_out/{chr}

selscan --xpnsl --vcf {pop}.{chr}.vcf.gz \
  --vcf-ref {ref}.{chr}.vcf.gz \
  --threads 10 --out xpnsl_out/{chr}

These calculations produce files with the extension .out containing raw iHS, XP-EHH or XP-nSL values. They are then normalized using the program norm. This performs a z-transformation across the genome for XP-EHH/XP-nSL scores and within each allele frequency bin for iHS scores. The resulting value in all cases is a z-score. It represents the number of standard deviations from the mean (in frequency bins for iHS).

To perform this normalisation the program, norm distributed with selscan was invoked as follows.

norm --ihs --files ihs_out/*.out --bins 50 --bp-win --min-snps 10 --winsize 50000
norm --xpehh --files xpehh_out/*.out --bp-win --min-snps 10 --winsize 50000
norm --xpnsl --files xpnsl_out/*.out --bp-win --min-snps 10 --winsize 50000

The outputs from norm are provided in two types of files. One set of files has extension .norm. These are normalised values for each variant position. Also provided are files with extension .windows which summarise the normalised statistics in non-overlapping windows of 50kbp.

Visualisation of scores genome-wide

Scores for all EHH based scans can be visualized across the genome using the pseudo-chromosome level coordinates. Translation to these coordinates was performed as follows;

python translate_coords.py {pop}.ihs.50bins.norm ragtag.scaffolds.agp > {pop}.ihs.50bins.norm.ragtag.txt 

Genome wide distribution of iHS scores

Since iHS is calculated for each population separately it is possible to directly compare scores between populations. The Manhattan plots below show several strong peaks in iHS distributed widely across the genome. Most of the strongest peaks in terms of maximum iHS values also seem to correspond with 50kb windows containing a high proportion of extreme iHS values. The strongest peaks appear to be both strong and wide, showing as dark or black dots due to overplotting of several extreme windows in the same region.

Figure 1: Manhattan plot of standardised |iHS| across the genome in three populations. Semi-transparent black points show the locations of the top 1% of 50kb windows based on the proportion of extreme |iHS| values within the window. Regions with many closely spaced extreme windows will appear darker

Genome-wide distribution of XP-EHH and XP-nSL statistics

Although the iHS statistic is sensitive to recent positive selection while accounting for local variation in recombination rate it loses power for sweeps that have resulted in fixation or near fixation of the selected allele. XP-EHH and XP-nSL statistics address this issue by comparing haplotypes in one population with those in a related population. Our three Acropora digitifera populations in WA are ideal for this because they diverged only recently and have similar demographic histories.

The following plots consider each focal population separately. Due to large number of loci, we only plot the loci with score > 2. Note that only positive scores are considered because we used combined haplotypes from both remaining populations as the reference. Positive scores in the XP* statistics will correspond to selection in the focal population only.

First we examine the inshore population, plotting all available EHH statistics. This seems to show considerable concordance between peaks in XP-EHH and XP-nSL but slightly less with iHS.

Figure 2: Manhattan plot of EHH based statistics (XP-EHH, XP-nSL, iHS) where the focal population is inshore.

The North Offshore population also has several very strong peaks but these appear to be mostly in different loci to inshore.

Figure 3: Manhattan plot of EHH based statistics (XP-EHH, XP-nSL, iHS) where the focal population is north offshore

For the South Offshore population we also see many strong sweeps including two very strong signals from the iHS statistic.

Figure 4: Manhattan plot of EHH based statistics (XP-EHH, XP-nSL, iHS) where the focal population is south offshore