Simple Question: What Is "Large" For Bwa?

Question

[Ketil]

I'm using bwa to align reads, and I have to choose the indexing method to use. The documentation says to use -a is for small genomes, and -a bwtsw for large genomes. I've used is, but sometimes this will crash with a segmentation fault, and bwtsw seems to work. Surely there is a better way to decide this than by trial and error? At what size is a genome "large", meaning I should use bwtsw? What is really the difference here?

Answer 1

[Vikas Bansal]

From BWA manual

 IS     is moderately fast, but does not work with database larger than 2GB.

bwtsw Algorithm implemented in BWT-SW. This method works with the whole human genome, but it does not work with database smaller than 10MB and it is usually slower than IS.

Answer 2

[matted]

It's not documented online as far as I can tell, but bwa 0.6.2 will choose the correct method automatically if you don't specify an algorithm.

The command line help hints at this:

Usage:   bwa index [-a bwtsw|is] [-c] <in.fasta>

Options: -a STR    BWT construction algorithm: bwtsw or is [auto]

Looking in the code, the decision is made by looking at the size of the reference genome:

// simplified sample from bwtindex.c:
if (algo_type == "auto") {
    if (l_pac > 50000000)
        algo_type = "bwtsw";
    else
        algo_type = "is";
}

I believe l_pac is the total number of bases in the reference (but the code is pretty dense so I'm not positive).