While I've been waiting for the Trimmomatic's eventual release, I took it upon myself to crank out my own perl script that seems to do much of what the Trimmomatic promises (except for adaptor screening). My question for tonybolger is essentially this: how do you run the sliding window? The way I've got mine working right now is it starts from the back end of the read and averages the quality of the last 4 bases (say bases 97, 98, 99, and 100). If this average quality is below the threshold, it clips the last base (100), the window moves backward by one base, averages the quality of the next 4 bases (96, 97, 98, 99) until the average of four bases is above the threshold. It then does the same sliding window deal from the front end.
Does this sound similar to what you're doing? Would it make sense to slide the window up by half its size each time (i.e. clip bases 99 and 100, then average bases 95 - 98), or even by a full window size?
Jes' wonderin' how you (or others) have approached it.
- A
P.S. Even in its current half-baked form, my trimmer script resulted in significant decreases in my contig numbers and increases in my max contig size and N50s doing de novo assembly in Ray (with SSPACE) and improved, but less spectacular numbers in Velvet. I think I'd been working under the illusion that quality trimming was more of a factor in aligning, but that coverage-based assemblers weren't as perturbed by low quality bases so long as coverage was deep enough (which mine has been). It's certainly looking like I was wrong about that...
Does this sound similar to what you're doing? Would it make sense to slide the window up by half its size each time (i.e. clip bases 99 and 100, then average bases 95 - 98), or even by a full window size?
Jes' wonderin' how you (or others) have approached it.
- A
P.S. Even in its current half-baked form, my trimmer script resulted in significant decreases in my contig numbers and increases in my max contig size and N50s doing de novo assembly in Ray (with SSPACE) and improved, but less spectacular numbers in Velvet. I think I'd been working under the illusion that quality trimming was more of a factor in aligning, but that coverage-based assemblers weren't as perturbed by low quality bases so long as coverage was deep enough (which mine has been). It's certainly looking like I was wrong about that...
Comment