Comments on: Expected Error Histograms from Illumina 36bp Reads http://lingpipe-blog.com/2010/05/10/expected-error-histograms-from-illumina-36bp-reads/ Natural Language Processing and Text Analytics Sun, 28 Apr 2013 22:30:55 +0000 hourly 1 http://wordpress.com/ By: lingpipe http://lingpipe-blog.com/2010/05/10/expected-error-histograms-from-illumina-36bp-reads/#comment-7097 Mon, 14 Jun 2010 17:31:55 +0000 http://lingpipe-blog.com/?p=3980#comment-7097 Thanks for the clarification. I’ll link to the comment from the erroneous part of the blog post itself.

Here’s a link to Bowtie’s doc for their --n alignment mode. It uses the sum of the quality scores at the mismatched base positions (quality here is probability of error on on a Phred scale), requiring it to be below a given threshold, and giving priority to better scores in cases of conflicts. They note that the backtracking size needs to be high in order to guarantee correctness (i.e. not make search mistakes).

I guess you could call this “quality aware”, but it’s not the kind of thing I was thinking about, which would actually use all the base probabilities in scoring the whole alignment. Like, say, GnuMap does.

]]> By: Mikael http://lingpipe-blog.com/2010/05/10/expected-error-histograms-from-illumina-36bp-reads/#comment-7089 Sun, 13 Jun 2010 21:55:16 +0000 http://lingpipe-blog.com/?p=3980#comment-7089 Actually, Bowtie does use quality scores, although you can set the program options so that it doesn’t. The default setting for Bowtie is to allow a certain number of mismatches in a “seed” region and, given that the read passed that first criterion, allow mismatches as long as their combined quality score does not pass a certain threshold.

]]>