bfast localalign -U option?

[Protaeus]

Hello,

What are the downsides of using the -U option at the bfast localalign step? I presume it'll be more computationally expensive, but are there things with the resulting data that would be of concern?

[nilshomer]

Hello,

What are the downsides of using the -U option at the bfast localalign step? I presume it'll be more computationally expensive, but are there things with the resulting data that would be of concern?

When not using the '-U', bfast enforces that where the seeds/masks match the reference during 'bfast match' also match the reference in the final alignment. This is a constrained Smith Waterman alignment, which speeds up the local alignment. Using the '-U' option ignores those constraints.

The philosophical question that the '-U' option lets one ask is if we get to a 100bp window in a 3Gb genome, should we constrain the alignment based on how we got there?

[Protaeus]

That brings up a separate but related question I had. A lot of my recent illumina runs have been 101bp on bacterial genomes. Generally speaking, the key width for my indexes is 30. Would there be any reason to believe I would benefit from trying a larger key width? Or is this where the -U option in localalign might be helpful?