I was taking a look at fastp the other day, and they have a rather interesting adapter trimming method. Cutadapt currently aligns the adapter sequence:
The advantage is thus that even very small adapter remnants can be removed and that this method is less sensitive to errors in the adapter sequence when adapter_length < 10 (with the 0.1 allowed errors cutadapt uses by default). Another advantage is that it only requires one pairwise alignment per read-pair rather than two.
The disadvantages are that since the alignment sequences are not known beforehand this limits opportunities to optimize.
Hi Marcel,
I was taking a look at fastp the other day, and they have a rather interesting adapter trimming method. Cutadapt currently aligns the adapter sequence:
But therefore it cannot detect:
Fastp uses its paired-end information for overlap analysis. Simply take the reverse complement and see where it matches.
As a result it can also match this:
The advantage is thus that even very small adapter remnants can be removed and that this method is less sensitive to errors in the adapter sequence when adapter_length < 10 (with the 0.1 allowed errors cutadapt uses by default). Another advantage is that it only requires one pairwise alignment per read-pair rather than two. The disadvantages are that since the alignment sequences are not known beforehand this limits opportunities to optimize.