when should I disable seeding in BWA?

[jesstilla]

Hello, I have previously been using BWA with the seeding disabled (by setting the seed length to be greater than the length of my reads). I originally started doing this because I was having trouble getting my reads to align in to some parts of the reference. However, I am now working with a different data set, and I am wondering whether disabling seeding still makes sense, given the increase in computational time.

So, I am wondering, as a general principle, under what conditions does it makes sense to disable seeding or not? Does it make sense to start with a seed in the first attempt at developing a pipeline and only turn to disabling seeding later if need be?

Thanks for your insight!