unmapped vs soft-clipping

[CNVboy]

quite confused about this two terminology. I'm reading Pindel, the split-read algorithm. The author seems to make use of the information of "unmapped" reads. Also there are other split-read-based algorithm, which uses "soft-clipped" reads, which are the unaligned parts of reads.

In my eyes, the two look quite similar. Say we have a 100bp read, 50bp of which cannot map while the 50bp can. Then how would BWA categorize this read? Will BWA think this is "unmapped" read since 50bp cannot be mapped; or it's "mapped" but with 50bp "soft-clipped" sequences?

Or BWA has a scoring system for mapping, which sets a threshold for distinguishing the two?

thx

[Heisman]

I haven't used BWA, but if it is "softclipped" (has an "S" in the CIGAR string) then I'd imagine that guarantees it is mapped (otherwise softclipping doesn't make sense to me). So, it would not be unmapped.

It would be worth checking if the latest version of Pindel will take in softclipped reads. Otherwise, you may want to extract them and feed them into Pindel as unmapped, somehow.

[bw.]

I also had this question earlier. This post cleared some things up:

http://seqanswers.com/forums/showthread.php?t=16842