tsinfer verify throw an error: ValueError: Alleles for sample 6 not equal at site 4

[chlangley]

Hello:
I have stumbled around this early impasse long enough. Time for a clue, please.

While I have gotten my data into a valid Sample and been able to add metadata (sampleIDs and populationIDs) and have run tsinfer infer on both the entire chromosome and on a small sub regions, test_direct.samples (see attached tsinfer_infer_out.txt. Note the unnerving line 13 in 'tsinfer_infer_out.txt' where the progress begins unexpectedly: "... sites; 7301 mutations######## ...".

When I examine indivdual trees on a simplify(d) subsample they appear to fit my expectations from previous analyses of these data.

Yet the resulting test_direct.trees does not tsinfer verify with test_direct.samples (see tsinfer_verify_out.txt . This seems like a good reason to pause and find out the meaning of the deviant genotypes.

The input file looks like this in the upper lefthand corner (snippet) where one might expect the " Alleles for sample 6 not equal at site 4" to be.
| => less ../cpragts/kgp4_chrX_hap_ba_snp
2781487 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781514 0 0 1 0 1 1 0 0 0 1 0 0 0 0 1 0 0
2781518 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781584 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
2781603 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781604 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781623 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781635 1 0 0 0 0 0 0 0 1 0 1 0 1 0 0 1 0
2781642 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781670 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781698 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781865 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781878 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781893 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2781927 0 1 1 1 1 0 1 0 0 0 0 1 0 1 0 0 1
2781983 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

I don't see an obvious candidate. Everything is 0,1,2 or -1.
How might one proceed to determine the source of this error?

thanks for any help.
Cheers,
Chuck

[molpopgen]

I think this belongs over at tsinfer, but I don't have permission to transfer. @benjeffery?

[jeromekelleher]

Done

[jeromekelleher]

Thanks for reporting this @chlangley. I haven't looked (sorry, short on time) but I suspect that the verify command doesn't deal with missing data. If I recall, it's just checking the input genotypes against the sample data, and it may not take missing data and imputation into account.

[chlangley]

OK, thanks. So the advise is to not attempt to use tsinfer verify on a SampleData with missing items (-1)... and forge ahead?

[jeromekelleher]

Yes, that would be my immediate reaction. Sorry, I don't have time to look at it properly right now...

This is a bug - as a minimum we should raise an error in this case saying this config isn't supported.

[jeromekelleher]

I've just verified that this is a bug @chlangley - I've opened #625 to track. Thanks for reporting this!