BWA: mismatch setting (Does '-n 3' allow up to three mismathces?)

[wwmm933]

Hi everyone,

I used BWA to map my paired end sequencing data. The length of read is 74bp. I would like to allow up to 3 mismatches in one read (no seed needed). I just set "bwa aln -n 3". Is this the correct setting for my purpose?

When I checked the mapped sam file (tag MD:Z), I saw more mismatches than three. So I am confused about mismatch settings in BWA. Can anybody help me?

Thanks!

[nilshomer]

Hi everyone,

I used BWA to map my paired end sequencing data. The length of read is 74bp. I would like to allow up to 3 mismatches in one read (no seed needed). I just set "bwa aln -n 3". Is this the correct setting for my purpose?

When I checked the mapped sam file (tag MD:Z), I saw more mismatches than three. So I am confused about mismatch settings in BWA. Can anybody help me?

Thanks!

BWA with "-n 3" will look for mappings up to four (to guarantee all up to three). Could you give an example record with more than four differences when using "-n 3"?

[wwmm933]

BWA with "-n 3" will look for mappings up to four (to guarantee all up to three). Could you give an example record with more than four differences when using "-n 3"?

Thanks, nilshomer! I just found the problem might not be the mismatch setting. I think it is due to paired end sequencing, so I got more mismatches on one end of paired reads. Is my understanding correct?

[wwmm933]

I have checked single end reads mapping. If I set '-n 3', the number of mismatches are well controlled. So the problem is that I used paired end reads.

[wwmm933]

Problem solved! Thanks!

[wanguan2000]

in bwa:
XM Number of mismatches in the alignment
MD Mismatching positions/bases

[alexbmp]

Well, I seem to have the same problem.

I ran bwa for paired-end reads, and no matter what value I put in -n option
( I tried -n 3, -n 4, -n 7, -n 0, ... even -n 0.05 ),

bwa finds more mismatches than I told it to do.

And what's worse; when I blat the mismatch-containing reads, their seldom in the accurate position!

-------

The above are only the facts.

It seems to me, ("seems" to me!) that the -n option restricts the # of mismatches for only one of the paired-end reads.

When I set -n 0, I found no paired-end reads that both contain mismatches: no pairs with both XM:i:# above 0, though one read has XM:i:0 and the other can have XM:i:# as high as 21, in the small example file I saw.

It's very confusing.

Isn't there any options to restrict both reads to behave the same?

I hope somebody sees this thread.

Have a great day!