BWA Combined Reference vs sequential alignment

[Tomi]

Hi,

I am currently developing a pipeline for novel viral detection and I am using BWA for filtering the mapped sequnces.

I recognized during my tests that I have a different number of unmapped reads after I align my sample against the whole genome once, in contrast to if I align each chromosome seperately and pipe the output.

Is this causing BWA heuristically concept, that I get a different number of mapped sequences by using the same references? Is this normal?

What is better, to align chromosome by chromosome or to align the whole genome at once?

Would be pelased if anyone could help.

[swbarnes2]

If you align against each chromosome individually, you will have some reads that ought to align to a different chromosome being forced to align to the chromosome you gave it. So the same read will align to multiple chromosomes, and would be counted multiple times. Does that explain what you are seeing?

It's definately better to give the aligner the whole reference sequence, so the aligner can find the best match, the real match for each read, rather than the best match of the sequences you gave it. Sure, it takes more memory, and more time, but that's what it takes.