Phased VCF data

[yashacode]

Hello,
I am new to proteogenomics. I am interested in using an unphased vcf file to make a protein fasta. Before I try out your tool, I was wondering how you handle multiple unphased heterozygous mutations per transcript. Will it include all mutations in a single translated protein, or will it include all combinations of the mutations in multiple translated products?

[ATPs]

I am still working on it. There are two options. One is neglect all samples and only use the first four columns. All variants were included in the transcript for protein translation.
One mode is like prohap, the same as generating protein for each person.

For vcf, unphased, it will be treated as phased. Two separate tsv files will be used.

There is a option to only use the variants and do not use the genotype information

[ddzhang3]

I have the same concern. For the unphased vcf, I apply genotype phasing tool, like shapeit to phase it using the 1000 genome reference panel. However, a proption of variants (about ~20% of all) are excluded in the phased vcf, as shapeit only consider the variants in the reference panel. So the PrecisionProDB-built customed proteome will miss the ~20% variants using the phased vcf. Do you have any suggestions on this isssue?

[ATPs]

sorry, I am not sure if I understand your concern correctly. I have updated PrecisionProDB so that now, it will first build a tsv file from the vcf file. If there are multiple samples to use, each of the sample will have two phased columns in the tsv file. The new protein models will be generated for each column. By default, the genotype like "0/1", they will be treated as if they were phased. I mean, 0 in the first column and 1 in the second column for a sample.

So if you do not want to miss any possible variants, you may generate the TSV file by your self. Include any variants you would like to include. The TSV file is very simple. Just put 0 or 1 in the sample column.

[ddzhang3]

I apologize for any lack of clarity in my previous explanation. My intended point is:
When processing genotypes at all heterozygous sites, this software assigns allele '0' to the first chromosome (e.g., paternal chromosome) and allele '1' to the other (maternal chromosome). However, in reality, the parental origin could be reversed—allele '0' might originate from the maternal chromosome, while '1' comes from the paternal chromosome.

The SHAPEIT software can phase ungenotyped heterozygous sites into paternal/maternal haplotypes to generate a phased VCF. Yet during this process, ~20% of variants are excluded because they are not present in its reference panel.

This creates a dilemma for downstream analysis:

If I use PrecisionProDB with the phased VCF, ~20% of variants will be missed.

If I use the unphased VCF, mis-phasing errors will propagate, leading to inaccurate personalized protein databases.

Could you share your perspective on resolving this technical conflict?

[ATPs]

Thank you for clarifying this. I guess it is out of the scope of what PrecisionProDB can do.

My suggestion will be, like I mentioned previously, you may create a tsv file manually. The vcf2mutation.py can run separately. You may generate a tsv file first and then use it as a input.

For phased VCF, you may added back the missed variants manually. Some coding skills are needed.

If you use the unphased VCF, you may generate the tsv file first, and then change the contents of sample__1, sample__2 manually and change them to "phased" distribution based on some evidence. But again, we cannot know which variant is from paternal and which variant is from maternal with the vcf file itself. Some "evidence" is needed. If with vcf from the father and mother, it is possible to know the origin?

But typically a gene will not get so much variants that will affect the proteins. If a gene has multiple variants, most variants will cause amino acid substitutions. If we use the proteins as a database for for mass spectremetry database searching, the chance of the variants affecting the same tryptic peptide is pretty low. So it should be OK in most case. Once you identified something interesting, you may take a closer look to see what is going on.