Weird output, only nan

[kellybarr]

Finally ran the program successfully by closely following the example. It ran with the following output:

==> Analysis will be run in 14365 combinations ==> GZIP input file (never BINARY) ==> Reading labels ==> Reading genotype data ==> Analyzing full dataset... ==> Freeing memory... Done!

Then the resulting file looks like this:

sample_name 0.0000000000 -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan -nan

That's just the first lines, but they all look like this. Anyone have any clue why this is happening?

[fgvieira]

can you send the full command line?

[kellybarr]

ngsDist --verbose 1 --geno BUOW.Plate1-3.rmrelFinal.baq.maf.glf4.glf --probs --n_sites 2568136 --n_ind 170 --labels BUOW.Plate1-3.rmrelFinal.ind.bamlist --n_threads 8 --out all_birds

[fgvieira]

can you send me some sample input files (maybe just 10-20 sites)?

[kellybarr]

sample.names.zip sample.glf.zip

[fgvieira]

How did you generate the GLF? With angsd? If so, what command did you use?

[kellybarr]

angsd -bam "$j" -P 8 -ref BUOW.fasta \ -uniqueOnly 1 -remove_bads 1 -baq 1 -C 50 \ -trim 0 -maxDepth 500 -doGlf 4 \ -minMapQ 20 -minQ 30 -minInd 40 -doCounts 1 \ -GL 1 -doMajorMinor 1 -doMaf 1 -skipTriallelic 1 \ -SNP_pval 1e-6 -doGeno 32 -doPost 1 -minMaf 0.05 -out "$outfile".maf.glf4

[fgvieira]

You are generating a files with 10 GL per site, but ngsDist needs as input only 3; try using -doGlf 3.

[TeresaPegan]

Hi, just to clarify, you are saying that -doGLf 3 is the only viable -doGLf option when using ANGSD to create the input for this program? I am confused because ngsLD takes files made with -doGLf 2 (e.g. https://github.com/fgvieira/ngsLD/issues/1), and the descriptions of input for ngsLD and ngsDist is almost identical and do not mention a reason why -doGLf 2 should work in one and not the other?

ngsDist:

Input data As input, ngsDist accepts both genotypes, genotype likelihoods (GL) or genotype posterior probabilities (GP). Genotypes >must be input as gziped TSV with one row per site and one column per individual n_sites.n_ind and genotypes coded as [-1, >0, 1, 2]. The file can have a header and an arbitrary number of columns preceeding the actual data (that will all be ignored), >much like the Beagle file format (link). As for GL and GP, ngsDist accepts both gzipd TSV and binary formats, but with 3 >columns per individual 3.n_sites.n_ind and, in the case of binary, the GL/GP coded as doubles.

ngsLD:

As input, ngsLD accepts both genotypes, genotype likelihoods (GP) or genotype posterior probabilities (GP). Genotypes >must be input as gziped TSV with one row per site and one column per individual n_sites.n_ind and genotypes coded as [-1, >0, 1, 2]. As for GL and GP, ngsLD accepts both gzipd TSV and binary formats, but with 3 columns per individual >3.n_sites.n_ind and, in the case of binary, the GL/GP coded as doubles.

[fgvieira]

both ngsLD and ngsDist accept as input files generated with angsd -doGlf option 2 or 3, but not 4 (which is the one you used).

[TeresaPegan]

Thanks. Actually it was a different poster using option 4, I had been trying with option 2 (beagle file) and it was telling me that my geno file had the wrong number of sites (even though I the number of sites was correct according to the number of lines in the unzipped beagle file), so I wondered if it just couldn't take the beagle format. I still have not gotten it to work with a beagle file (produced by ANGSD's -doGLf 2) because of this error about the wrong number of sites, but that's really a separate issue.

[fgvieira]

When you counted the lines of the beagle file, did you subtract the header? If so, can you then send me the beagle file you are using?

[TeresaPegan]

Ahhh, counting the header was the problem for me. Thanks!!

[fgvieira]

No problem... :smile: I'll close it now, but feel free to reopen if the problem persists.

[TeresaPegan]

Today I observed that I get this weird output, only nan, when I run ngsDist with --tot_sites 1, as discussed in this issue. However, when I run the same code without the --tot_sites option, I get normal output. I can see that is not what the original poster was doing, but decided to comment here in case it's helpful to know this, since the effect was the same -- a matrix of nan's.

[fgvieira]

Thanks Teresa for reporting the issue.

Indeed there was an issue that has now been corrected. Just specify --evol_model 0 to get the raw p-distance (by defalt ngsDist outputs a log transformed p-distance)

Thansk,