2d SFS Estimation

Angsd can estimate a 2d site frequency spectrum. This is an extension of the 1d site frequency spectrum method.

• Newer versions of ANGSD can estimate even higher dimensions (upto 4).
• From august17 2019 the program can now do a proper folding of the 2dsfs, which is done by supplying it with the UNFOLDED saf.idx fiels generated by -dosaf 1

Below are some examples: And is best explained by a full example.

Example

• Assume you have a 12 bamfiles for population in the file pop1.list
• Assume you have a 14 bamfiles for population in the file pop2.list
• Assume you have a fastafile containing the ancestral state in the anc.fa

Let's start by finding the positions for which we have data in population1 and population2

# as always you can add -minMapQ 1 and -minQ 20 to only keep high quality data.
angsd -GL 1 -b pop1.list -anc anc.fa -r chr1: -P 10 -out pop1 -doSaf 1
angsd -GL 1 -b pop2.list -anc anc.fa -r chr1: -P 10 -out pop2 -doSaf 1

1 dimensional frequency spectra

If we were interested in estimating the 1d sfs for each population we could do it like this using the realSFS program. (See more on page )

#sfs for pop1
realSFS pop1.saf.idx -P 24 >pop1.saf.sfs
#sfs for pop2
realSFS pop2.saf.idx -P 24 >pop2.saf.sfs
#2d sfs for pop1 and pop2
realSFS pop1.saf.idx pop2.saf.idx -P 24 >2dsfs.sfs

The output is then located in a nice flattened matrix format(25x29) in the file: 2dsfs.sfs. Good luck visualising it, some people are using dadi, we have been using heat maps in R.

2d sfs (folded)

#2d sfs for pop1 and pop2 doing proper folding
realSFS pop1.saf.idx pop2.saf.idx -P 24 -fold 1 >2dsfs.sfs