Invariant sites not included in no_sites

[gfwalker]

Hello,

I am trying to use pixy on data generated using a reduced representation RAD protocol. I assembled loci using the Stacks program (Catchen et al. 2013) and generated a vcf including both variant and invariant sites (data_subset.vcf.gz) with the Stacks Populations module. However, when I run pixy on the dataset, the results indicate that only the variant loci are considered valid and used for calculating pi. In the attached data and output, chromosome (RAD locus) 4 contains 303 total sites, 7 of which are variant. However, in the pixy_pi_subset.txt output file, you can see that no_sites = 7, suggesting that only 7 sites "have at least one valid genotype", according to the Pixy documentation. What is causing my invariant sites be considered invalid genotypes?

Code used to run Pixy:
pixy --stats pi fst dxy --vcf data_subset.vcf.gz --populations popmap.txt --window_size 10000 --n_cores 4

data_subset.vcf.gz (note: needs to be indexed prior to running as .tbi file would not attach)
popmap.txt
pixy_pi.txt

Thank you,
Geoffrey

[ksamuk]

Hi there, can you try rerunning this with version 1.2.5? You can install it by: conda install -c conda-forge pixy=1.2.5

[gfwalker]

Yes, thank you, that worked! I had seen that fix on some other threads, but it seemed to apply to a different issue. Thank you!

For anyone else who runs into the same issue, I am on an Apple Silicon Mac, so I needed to create a conda environment with --platform osx64 to install the older Pixy v1.2.5.

[ksamuk]

Good stuff! I think this was related to the workaround for the new GATK missing specification (that has now been reverted by the GATK team due to overwhelming opposition). Either way, its a bug and we should have a fix in the new major update, but the numbers will be the same.