Not primary vs Supplementary alignment

[dnusol]

Hi,

can anyone explain to me the difference (if there is one) between a read having a 0x100 flag (not primary alignment) and a read having a 0x800 flag (supplementary alignment)? How does the alignment differ between these two reads? can one read have both flags assigned?

Cheers,

Dave

[dpryan]

The 0x800 concept is relatively new and meant to convey chimeric/non-linear alignments. Suppose you have a chromosomal inversion with a read mapping over its junction point. The first part of the read will map normally and then be truncated. The second part of the read will map further up/downstream and in the wrong orientation. Because of this, it doesn't map linearly along the chromosome so the 0x800 bit in the FLAG is set.

I should note that the lack of the 0x800 bit being set doesn't always mean that a read is not chimeric. Tophat2, for example, can produce mappings, but doesn't set that bit (tophat-fusion predates the addition of the 0x800 bit to the SAM spec.).

[dnusol]

Thanks for your input dpryan, it is very helpful.
I think I understand the example of a chromosomal inversion. I am trying to interpret the SAM format for such reads.
So does a chimeric alignment cut the read in two? and thus in the SAM format a chimeric alignment will have a primary alignment with a CIGAR of lenght say 'X' less than the total read length, and a SA: tag with a length say 'Y' less than the total read lenght?
Is this the difference from a read indicating that the alignment is not primary?

[dpryan]

Yes, chimeric alignments cut reads in two (or more). If the read was originally 100bp, then the first alignment might just be 40bp and the second 60. Usually the entirety of the sequence isn't put in either of the alignments (i.e., it's not hard-clipped). A secondary alignment will be of the same length and sequence as the primary alignment, since it's just an alternative mapping (i.e., the read is a multimapper).