Tool that can merge 2 VCF files while taking "representational ambiguity" of (multi-allelic) variants into account

Question

[William]

Is there a tool that can merge 2 VCF files while taking "representational ambiguity" of multi-allelic variants into account?

By:

• replaying all variant alleles from the 2 VCF files into the reference genome
• identifying which alleles are actually the same but just written down in a different way
• calculating what the best way is to represent the merged variants/alleles in a new (multi-allelic) variant

See also this question and answer. Should you decompose and normalize multi-allelic variants for comparison / ID assignment?

The (multi-allelic) variants (alleles) in both VCF files are different because:

1. different technology used to make the VCF files
2. different alternative alleles present in samples

BCFtools merge does not take "representational ambiguity" of variants into account (as far as I know)

First decomposing and normalizing all variants to bi-allelic in both input VCF files, then merging and collapsing overlapping variants back to multi-allelic destroys some information?

[Ram]

First decomposing and normalizing all variants to bi-allelic in both input VCF files, then merging and collapsing overlapping variants back to multi-allelic destroys some information?

Can you give me an example? Your three point thing up top is a description of the left-aligned most parsimonious representation done by vt normalize/bcftools norm (I prefer the former)