SAM/CIGAR for deletion before first match

[stuartrbaker]

Hi all,
This is my first post here. Just searching this site has answered so many questions I've had previously, but I can't find an answer to this one. If I have hard trimmed some bases from the beginning of my reads after alignment, this leaves me with the possibility of a correctly-mapped deletion beginning a read. The SAM spec seems ambiguous to me on this topic, and if this is valid, which position should I take to be the leftmost mapped position? For example, say I hard trim the first three bases from the following aligned read:

reference: ACTGACGTCA
read: ACTCGTCA
CIGAR: 3M2D5M

This is now:

read: CGTCA
CIGAR 3H2D5M

There may not be a correct answer to this question, so maybe I'm just asking will SAMTOOLS pileup get confused by this?
Thanks!

Stuart

[Heisman]

I think if you hardclip the first 3 then the CIGAR shouldn't show any deletion; you would be aligning the read "CGTCA", which directly matches the reference.

[stuartrbaker]

Thank you for your reply. While I can opt out of this behavior from the aligner, I don't want to do this because very much want to retain the deletion. In these cases the deletion is correctly identified and aligned, and actually more "interesting" data than a match. I have written my software to deal with a leading deletion, but I am just curious about whether there is an official usage with SAM.

[maubp]

I think your usage looks sane - it would be worth bringing this up on the samtools-devel mailing list. I suspect similar examples could occur in assembly editors, e.g. Gap5 might generate this kind of CIGAR string too after a manual edit.

[jkbonfield]

I'm not sure what Gap5 will do without checking, but it may depend on whether the usual has manually added gaps with * characters.

I think starting in a deletion is fine, although it's hard to justify the existance of it instead of a positional change. However given the deleted base does have its own location in the reference I would expect samtools to work, although I haven't verified it.

Where things start to break down is with sequences that start with an insertion or padding character. This can be a useful and valid alignment, but it causes confusion to samtools as the first used base occurs before the mapping position.