mpileup VCF multiple base reference indels

[cram]

I have a question about a quirk of the VCF output by samtools (0.1.18) mpileup.

In some cases insertions are reported with two or more bases in the reference sequence, for example:

Chr1   1790    .   CA  CAA 117 .   INDEL;...

It was my impression that the reference sequence existed only to say "the insertion happened after this point", and so I don't understand why it's necessary to include any preceding bases, the C in this case, rather than simply reporting it as:

Chr1   1791    .   A   AA  117 .   INDEL;...

Is there some significance to this that I am missing?

[krawitz]

There is now the "left align" convention in reporting indels. In vcf you describe an insertion by reporting the last unchanged base you from the left that is for sure. Let's assume your reference sequence is caag and your samples sequence is caaag. There are basically three possibilities:
chr1 1 . C CA
chr1 2 . A AA
chr1 3 . A AA
However, with the left align convention only the first one is correct.
We discussed the problem here http://www.ncbi.nlm.nih.gov/pubmed/20144947. We made another suggestion for reporting indels in an unambiguous way. But the left align rule is now the de facto standard, which is fine, too.