Closed LauraVP1994 closed 10 months ago
Hey @LauraVP1994! It sounds like you may be running into an issue where there's basically zero sites with your specified sequencing depth. @dylanpilz, can you take a closer look?
Hey, as @joshuailevy pointed out, none of the sites in the sample have a higher coverage than 500, so all of the available data has been excluded. I think the error messaging might be partly to blame, since it always suggests increasing --depthcutoff
whenever a solver error occurs. I'll add a check that determines whether the depthcutoff is set too high for the depth available in the sample.
Using Freyja version 1.4.7 with barcodes from 12-08-2023, I was only able to get a warning rather than an outright error, so it's quite possible that the version of usher_barcodes.csv you're using is incompatible with this sample. Could you try running freyja update
and see if the issue persists?
Thank you, it is indeed probably because of the low coverage of the sample. We use freyja as part of a larger snakemake script that should analyze samples from beginning to end, so it threw errors because no file was generated as a consequence, but we have solved this by creating an empty file if this error occurs.
I also had another question, the results show which variant from the barcode options is the closest, but is there also a way to know how close to the match the results are?
Sounds great!
The resid
field in the demix output gives the residual of the demixing problem, which describes how far the variants in the barcode file are from those observed in the sample at the provided depths.
Dear,
Thank you for the answer, I gues this is the resid in the demix file? What is the range? When can you assume that it is a good resid?
Kind regards Laura
Correct, you'll find resid
listed in the demix file.
As to whether a given resid
is considred good, that's going to depend on the type of sequencing data you're looking at. For Illumina sequencing of SARS-CoV-2, 20 or below is generally considered good, but for ONT sequencing that value could be signifcantly higher due to it being more error-prone.
At it's core, it's going to depend on how well your available sequencing data is able to capture viral diversity.
Hello,
I have been using freyja and there are some errors that keep persisting due to "--depthcutoff"? It also seems that like the type of error ouptut is related to the version of freyja. It should be noted that this problem is especially in samples with low coverage:
Freyja version 1.4.5 Input code:
freyja demix sample_ivar_freyja.variants.tsv sample_ivar_freyja.depth --eps 0.025 --depthcutoff 500 --output sample_ivar_freyja.demix
Error
Input code:
freyja demix sample_ivar_freyja.variants.tsv sample_ivar_freyja.depth --eps 0.025 --output sample_ivar_freyja.demix
Freyja version 1.4.7 Input code:
freyja demix sample_ivar_freyja.variants.tsv sample_ivar_freyja.depth --eps 0.025 --depthcutoff 500 --output sample_ivar_freyja.demix
Error
Input code:
freyja demix sample_ivar_freyja.variants.tsv sample_ivar_freyja.depth --eps 0.025 --output sample_ivar_freyja.demix
Error
sample_ivar_freyja.depth.txt sample_ivar_freyja.variants.tsv.txt
I also attached the files (.txt should be removed at end).
Thanks for any help!