View Single Post
Old 04-28-2014, 01:30 AM   #36
Location: Münster, Germany

Join Date: Mar 2013
Posts: 42

Originally Posted by Brian Bushnell View Post
If you use "random" then possibly (probably?) you shouldn't include the xmtag, as that might confuse cufflinks.
This in in accordance with my gut feeling, since I interpet the Cufflinks docs in the way, that the algorithm adjusts the read weight based on the xmtag:

By default, Cufflinks will uniformly divide each multi-mapped read to all of the positions it maps to. In other words, a read mapping to 10 positions will count as 10% of a read at each position.
However Cufflinks in a later stage also needs the secondary alignment positions for the fine grained abundance estimation:

Cufflinks will then re-estimate the abundances dividing each multi-mapped read probabalistically based on the initial abundance estimation of the genes it maps to
But thanks Brian for working this out!

Last edited by Thias; 04-28-2014 at 01:31 AM. Reason: Typo
Thias is offline   Reply With Quote