View Single Post
Old 03-05-2013, 12:26 AM   #2
Simon Andrews
Location: Babraham Inst, Cambridge, UK

Join Date: May 2009
Posts: 871

There are two reasons to trim, either to remove poor quality sequence (quality trimming) or to remove adapter sequence where your read length is long enough that it's read through your insert and into the adapter on the other end.

For long reads adapter trimming is very useful. Depending on the construction of your library it is possible to end up with a significant proportion of your library containing some adapter sequence. If you don't trim this then you will leave non-native sequence on the end of your reads which will cause all kinds of problems with subsequent mapping or assembly.

Quality trimming is not quite so clear. If you have a fully quality aware analysis pipeline then it's arguable that you should leave all base calls in and let the downstream analysis make what use out of them that it can. In reality many tools are not quality aware, or make a somewhat crude use of qualities such that you get better results from completely removing poor quality data. From a practical point of view you can also run your analysis more quickly if you initially filter your data to remove calls that you're pretty confident are never going to tell you anything useful.
simonandrews is offline   Reply With Quote