Why BWA output didn't agree with bowtie2

[jdpr_100]

I am working on the alignment between single-end read and reference genome using BWA and bowtie2. I got different results from BWA and bowtie. I couldn’t figure out what I am doing wrong.
bowtie2 commands:
bowtie2 --sensitive-local -x <indexed ref> -q <reads> -S output.sam
I got following statistics with bowtie2.
total reads 411526452
69195599 (16.81%) aligned 0 times
304002593 (73.87%) aligned exactly 1 time
38328260 (9.31%) aligned >1 times
I also check the XM tag in .sam file generated from bowtie2
31720802 XM:i:5
46327246 XM:i:4
58292573 XM:i:3
59969734 XM:i:2
46552300 XM:i:1
22355229 XM:i:0
When I run bwa with the following commands, I would like to generate similar statistics.
bwa index ref.fa
bwa aln –n 5 ref.fa reads.fq > out.sai
bwa samse ref.fa out.sai reads.fq > out.sam
And , get some stats with following commands.
samtools view -c output.sorted.bam
samtools view -c -q1 output.sorted.bam
samtools view -c -F 4 output.bam
samtools view -c -f 4 output.bam
total reads 411526452
# of reads uniquely mapped 179363510 43.58%
unmapped reads 230229393 55.94%
mapped reads 181297059 44.05%
I also check the XM tag in output.sam generated from bwa.
18855884 XM:i:5
32102066 XM:i:4
41821781 XM:i:3
43223159 XM:i:2
31594809 XM:i:1
13301580 XM:i:0
There are huge different between the results from BWA and that from Bowtie2. Only about 44% reads have been aligned with the ref using bwa.
What is the reason cause this difference?
Thank your for your help in advance

[dpryan]

Why would you expect them to be similar? Different algorithms will always produce different results.

[jdpr_100]

Compared to 71% in Bowtie2, only 44% reads can be mapped to reference genome using BWA. I expect at least 60%. I am not sure whether the results are normal or not. Or, there something wrong with -n 5 (mismatches) in BWA.

[Brian Bushnell]

Clearly, the problem here is that you're not using BBMap

But when using bwa, you should run bwa-mem instead of bwa aln, which is slower and cannot handle as many errors. Bwa-mem will give you output that is more similar to bowtie2.

[jdpr_100]

Hi Brian Bushnell, Thank you for your suggestion. In my case, reads length range from 50 to 100bp. Can bwa-mem handle 50-100bp reads?

[Brian Bushnell]

Yes, it can.