View Single Post
Old 06-03-2014, 11:44 PM   #59
WhatsOEver
Senior Member
 
Location: Germany

Join Date: Apr 2012
Posts: 215
Default

Sorry, but I don't get it completely.

Quote:
Originally Posted by Brian Bushnell View Post
I use the "XT:A:R" tag to indicate that a read is ambiguously mapped and XT:A:U to indicate unambiguously mapped. If you say "ambig=all" then the threshold for printing secondary alignments is more liberal than the threshold for deciding an alignment is ambiguous. So, even "unambiguously" aligned reads will sometimes have secondary sites printed - the best one, and others that are "decent" even though they are sufficiently worse than the best site that I classify the best site as unambiguous.
What is your definition of an "unambiguously mapped read" if it has secondary alignments? If secondary alignments are printed as a result of a more liberal threshold, the read is no longer unambiguous under the used threshold, is it? This should be independent of how bad your sec alignments are, the chosen threshold shouldn't change the terminology, should it?
So in my case I have ~16K reads which are "unambiguous" according to your definition but have seconday alignments?

Quote:
Originally Posted by Brian Bushnell View Post
So if you add up all of the primary alignments with XT:A:U, you should get the same numbers as the program reports.
You are right - if I count XT:A:R, I get the same number as the program output (XT:A:U are however not visible/set?, but should of course be the remaining)

Quote:
Originally Posted by Brian Bushnell View Post
Sorry that it's a custom tag (that is also used by bwa), but there's nothing in the official flag bits or official tags to indicate whether an alignment is ambiguous, and the SAM spec doesn't say anything about the presence of secondary alignments indicating ambiguity. It's really a subjective judgement.
Yes, I know that and I really think that this is a confusing point in the SAM spec. But no, I don't think that it's a subjective judgement just because the word "ambiguity" is not explicitly used in the specs.
Or is it just my german misinterpretation of the strictness in the word ambiguity?

But being aware of that also helps, thanks again Brian for a fast and helpful response. I will now stop my work on mapper-evaluation for my projects and focus on the next steps - so I won't bother you with additional things in the next time
WhatsOEver is offline   Reply With Quote