nothing aligned using bwa mem

[demienx]

Hi there.
I am using bwa to align human RNA-seq reads to rRNA database to exclude rRNA pollution. I tried both bwa backtrack and mem, and very strangely only backtrack could map some of my reads to rRNA reference, while mem mapped NONE. I think there must be something wrong but I just cannot figure out what. My command is listed below:
backtrack:
bwa aln -l 20 -f se-reads.sai rRNA.fa se-reads.gz && bwa samse -f se-read-backtrack.sam rRNA.fa se-reads.sai se-reads.gz
mem:
bwa mem -k 17 rRNA.fa se-reads.gz > se-read-mem.sam

My reads were single-ended and the length was 28 bp. The input reads were fastq gzip files.

And I'd like to post a same read for example that was mapped to rRNA reference using backtrack but was not mapped using mem.
backtrack:
GS85516-FS3:L02C002R005.7860.0 0 hsa_rRNA_EU597543_650:1603:+ 519 0 28M * 0 0 AAAGGACCTGGCGGTGCTTCATATCCCG ,,+/#*$#+$##$%,#)')0..1..)#) XT:A:R NM:i:1 X0:i:5519 XM:i:1 XO:i:0 XG:i:0 MD:Z:27T0
mem:
GS85516-FS3:L02C002R005.7860.0 4 * 0 0 * * 0 0 AAAGGACCTGGCGGTGCTTCATATCCCG ,,+/#*$#+$##$%,#)')0..1..)#)AS:i:0 XS:i:0

I am very confused. Can anyone help me with this?

[Ihavenoidea]

Hello,
I work with miRNA data and I had the same problem, bwa-mem was mapping 1-5% of my reads, but when I used backtrack, my mapping got up to 80-85%. Mem is optimized for longer reads and doesn't work well for short fragments.
Also, I have seen a lot of differences in the mapping depending on the specifity of my trimming, going from 60 to 80% with bwa-aln + samse.
Hope is of any help for u!

[HESmith]

From the BWA-MEM abstract:

"The algorithm is robust to sequencing errors and applicable to a wide range of sequence lengths from 70bp to a few megabases."

The length of your example is 28bp - too short for this tool.

[demienx]

Thank you Ihavenoidea. It seems very possible that the low mapping rate is due to my short read length. But could you tell me how long your miRNA read is?

[demienx]

Hi HESmith,
I should apologize for making such a simple mistake, and thank you for pointing it out for me. I originally thought that mem was suitable for a wider range of read length than backtrack.

[Ihavenoidea]

Thank you Ihavenoidea. It seems very possible that the low mapping rate is due to my short read length. But could you tell me how long your miRNA read is?

After efficiently trimming my samples are of about 20pbs