Error while running demix on VCFs produced by bcftools

[watronfire]

Hi all,

I ran into the following issue while running freyja demix on a VCF generate by bcftools:

building mix/depth matrices
Traceback (most recent call last):
  File "/PHShome/nm104/mambaforge/envs/phonebook/bin/freyja", line 10, in <module>
    sys.exit(cli())
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/click/core.py", line 1157, in __call__
    return self.main(*args, **kwargs)
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/click/core.py", line 1078, in main
    rv = self.invoke(ctx)
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/click/core.py", line 1688, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/click/core.py", line 1434, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/click/core.py", line 783, in invoke
    return __callback(*args, **kwargs)
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/freyja/_cli.py", line 124, in demix
    mix, depths_, cov = build_mix_and_depth_arrays(variants, depths, muts,
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/freyja/sample_deconv.py", line 64, in build_mix_and_depth_arrays
    df = read_snv_frequencies_vcf(fn, depthFn, muts)
  File "/PHShome/nm104/mambaforge/envs/phonebook/lib/python3.9/site-packages/freyja/sample_deconv.py", line 104, in read_snv_frequencies_vcf
    df["ALT_FREQ"] = vcf_info[j].str.split('=')\
ValueError: could not convert string to float: '0,0,1,6'

The specific command I used was: freyja demix OUG-1858.variants-filled.vcf OUG-1858.depth.txt --output test.txt --barcodes freyja-barcodes/usher_barcodes.feather

All the files are attached. freyja_demix_files.zip

The traceback points to the following code:

Freyja/freyja/sample_deconv.py

Lines 116 to 124 in af5f46a

	vcf_info = df['INFO'].str.split(';', expand=True)
	for j in range(vcf_info.shape[1]):
	if vcf_info[j].str.split('=')[0] is not None:
	if vcf_info[j].str.split('=')[0][0] == 'AF':
	df["ALT_FREQ"] = vcf_info[j].str.split('=')\
	.str[1]\
	.values\
	.astype(
	float)

Based on my debugging, it seems the function assumes that all variants have the same number of fields in the INFO column, but I guess VCFs produced by bcftools don't (indels have extra fields in my files). As a result, I was able to workaround the problem by just extracting the value of the AF field using regex.

Environment:

• python 3.9.19
• pandas 2.2.3
• usher 0.6.3
• freyja 1.5.2