Genome Coordinate Use and VRS

[mbaudis]

The current use of referenceBases and alternateBases is partially incorrect when compared to the GA4GH VRS standard. While it is fine for SNVs (i.e. replacement of a single reference base by another base) or compound replacements (e.g. A > GG), it is incorrect for simple INDELs, i.e. insertions or deletions of one or more bases w/o change of the surrounding sequence. Here, VRS states (in spirit) that

• a deletion results in the removal of the reference base(s) and an empty value for the replacement
• an insertion happens at a 0-based, 0-length interval w/ empty reference base and representation of the added bases.

This (preferable) use also removes some logical inconsistencies when using VCF-style "anchor" bases for INDEL variants, e.g. regarding the length of variants or their position.

However, this needs some implementation considerations:

• ReferenceBases.pattern should change from ^([ACGTN]+)$ to ^([ACGTN]+)?$ or ^([ACGTN]*)?$
• the same applies to AlternateBases
• this should also be propagated to front-end examples and shorthand annotations (e.g. ATC>A style should not be used)
• backends should consider alternative searches if containing "anchor-base" annotations
  • {start: 1234567, referenceBases: "ATC", alternateBases: "A" } is equal to
  • {start: 1234568, referenceBases: "TC", alternateBases: "" }

Discuss, please.

[mbaudis]

And gladly accept corrections / refinements @ahwagner ...

[sdelatorrep]

Hi @mbaudis. The specification part was changed in PR #41, wasn't it? If so, I suggest closing this.

[mbaudis]

Closing since done in https://github.com/ga4gh-beacon/specification-v2/pull/41.