That sounds perfect Daniel. Advanced options for advanced users, and safe defaults for everyone else will do it. Florent On 07/04/10 01:00, Daniel Blankenberg wrote:
Hi Florent,
Thanks very much for the comments. A sliding window sounds like an excellent approach: allow users to specify the window size, step size, an aggregation action to perform on the window (min, max, sum, mean, etc ), a comparison method (<,<=, ==, etc) and a threshold quality value; allowing users to specify the ends (both or only one or the other) to trim would also likely be useful. Would it also be desirable to allow specifying a number of quality scores that can be excluded from the aggregation action (the zero low quality base pairs in your example)? A window size of 1 would handle the simple case of only trimming the very ends while allowing the user to configure more complex windowing schemes. Thoughts?
Thanks,
Dan
On Apr 6, 2010, at 4:00 AM, Florent Angly wrote:
Thanks for your reply Daniel.
You are correct that there is not currently a tool to trim directly by quality in Galaxy; currently the the Summary statistics and boxplot tools are used to determine good cut off for use in the trim by column tool; percentage of read length can be more useful on variable length reads. However, adding a tool that can directly trim reads based upon a threshold quality score seems like a natural fit for Galaxy, when uniform read length is not present at the start and/or not a requirement at the end and the percentage-of-read-length method is not sufficient
That's right... I did not even think about using the boxplot tool to find how much to trim the ends. My reads all have the same length, but still, is seems more natural to only trim as much as needed and no more. For example, I have some reads that are completely low quality and should entirely trimmed/removed, whereas some might of good quality over almost all their length.
Lets verify that you are looking for something like this, where 'x' is a low quality base and 'o' is a high quality base: Start with: xxxooooxxooooxxx after trimming ends for 'x': ooooxxoooo So that trimming happens only from the ends and stops as soon as a base above the threshold is found and internal low quality bases are not considered.
It's probaby better to use a short sliding window (of, say, 5 bp) and trim the ends until the window has no more than, say zero low quality base pairs. So, the following sequence would be converted from: xxxoxooooooxxooooooxoxxx to: ooooooxxoooooo
Florent
_______________________________________________ galaxy-user mailing list galaxy-user@lists.bx.psu.edu http://lists.bx.psu.edu/listinfo/galaxy-user