Differences between the application of function end-to-end and local in Bowtie2

Question

[Abieskawa]

Hi everyone, I want to ask about the differences between the application of function end-to-end and local in Bowtie2. I have found that others said that they think local alignment does good for genotyping-by-sequencing(GBS), since it is quick and skips lots of mismatches & gaps (insertions/deletions) outside the region of similarity. Here are the resources: Resoucres 1Resources 2

I wonder if we should also apply the "local" function to align whole-genome data to the reference. If there is someone who can explain more about the application of these two functions, I am also welcome to that.

Answer 1

[Asaf]

Local alignment searches for the best alignment of a substring of the input sequence. While it can find an alignment for the entire sequence, if another, shorter, alignment has a higher score it will be chosen.

End-to-end will compute the score over the entire matching of the input sequence and its alignment with the reference.

Basically if there are adapters/long mismatches/indels etc. the local will work best. If you have a good reason to believe that the input sequence should be fully matched to the reference then select end-to-end.

[Abieskawa]

So...If I have already cut adapters from the sequence, I should choose end-to-end, right?

[Asaf]

Not necessarily. If you have an insertion near the end of the read then it might get aligned with local but not end-to-end

[ATpoint]

End-to-end is the default and appropriate for most standard assays such as ChIP-seq or ATAC-seq. For variant calling (for which many callers assume bwa mem anyway) you might need to change that. I would (in fact I do) leave it at default most of the time. Do that, see alignment percentage and then decide if it is worth changing it.