Dynamic read trimming of poor quality reads

[GoogleCodeExporter]

We have observed certain sets of reads that have poor quality bases at the 3' 
end.  This can cause 3' end segments to be unmapped, and place a greater burden 
on the segmented alignment phase.  Not only is this slower, but it will 
inevitably lead to a greater number of false positives (garbage in => garbage 
out).

Dynamic read trimming is a viable solution here.  The first goal is to follow 
the BWA trimming paradigm. 

Note that this feature cannot be enabled until variable length reads are fully 
supported throughout the pipeline.  There are still areas where fixed length 
reads are assumed and shortcuts are taken.

Original issue reported on code.google.com by [email protected] on 10 Feb 2011 at 4:20