bwa multi-alignment behaviour

[The Snow]

Hello everybody,

I'm quite newbie with bioinformatic, so please forgive me if this is a silly question.

Currently I'm align my reads (100bp, single end) against a .fasta containing a list of viruses (ref seq from NCBI). I'm using bwa (sampe)

If I align my reads against a fasta containg 1 virus I have the following results:

analysis1
826 reads aligned versus Virus X

If I align my reads against a viral database I have the following results:

analysis2
740 reads aligned versus Virus X
33 reads aligned versus Virux Y
50 reads aligned versus Virux Z
3 reads that are missing (very strange)

Note that I perform the analysis two time, I don't use the same bam/sam file for both alignment (of course).

So I was wondering why bwa is not working in this way:

826 aligned versus Virus X
AND
33 reads against Virus Y
50 reads against Virus Z

Otherwise with a multiple alignment I can miss some data (e.g. analysis2)....right?

Is there something that I'm missing?

Thank you!

Fabio

[swbarnes2]

Some of those 86 reads that used to align to virus X when it was the only available reference to align to align better to virus Y or Z than they did to virus X, so that's where they align when the aligner is given all three to choose from.

So you aren't missing data by aligning to multiple viruses, you are getting more accurate results.

[The Snow]

Yeah, that's what I've thought. The problem is related to the result: if I have 740 reads of virus X and 50 reads of virus Y, how I can assure that we are not in presence of two viruses but just one (Virus X)? Should I realign my reads (every time) vs each virus detected (one-by-one) in order to be on the "safe side"?

Furthermore in the alignment of the virus detected there are some region of its genome that are missing...in articular those related to repeated regions. Is there a way to manage them with bwa? I think that the repeated region (~800kb-1000kB) are difficult to be analyzed with bwa. Do you have any tips?