miRNA mapping with bowtie2 gives low alignment %

Question

[Ant]

Hello, I'm doing a miRNA-seq analysis but the alignment is extremely low and I can't figure out why. The QIAseq® miRNA UDI Library Kit was used. The wet lab staff said that only 3' adapter needs to be removed.

Before trimming the read length was 75bp and after trimming the read length was on average 23bp, so I don't know what could be wrong. the parameters used were these:

cutadapt -a AACTGTAGGCACCATCAAT -m 18 -M 34 \ -o "data/trimmed_reads/${base}.trimmed.fastq" "$file" \ data/trimmed_reads/${base}.cutadapt.log

bowtie2 --local -N 1 -L 16 -x data/alignments/mirbase -U "data/trimmed_reads/${base}.trimmed.fastq" -S "data/alignments/${base}.sam"

and in bowtie2: I used the mature.fa provided by miRBase for the alignment and changed the U-->T.

But the result doesn't change much from this: Analysis complete for XXXX.trimmed.fastq 85645 reads; of these: 85645 (100.00%) were unpaired; of these: 84956 (99.20%) aligned 0 times 0 (0.00%) aligned exactly 1 team 689 (0.80%) aligned >1 times 0.80% overall alignment rate

I already tried changing the parameters for both cutadapter and bowtie2 but nothing changed much. Maybe I need to cut something else?? Fastq indeed saw some PCR contaminants but the weblab staff said to be more rigorous in cut so it would disappear, I don't think I can be more than that??

I would appreciate any help.

[GenoMax]

Try bowtie v.1.x since you want ungapped alignments for miRNA.

[Ant]

okay, I will try and come to tell the results, thanks

[Ant]

Hi, I used bowtie1 1.3.1 and it increased a little, but it wasn't very satisfactory.

I used the parameters:

bowtie -n 0 -l 15 -e 99999 -k 200 --best --strata -x data/alignments/mirbase "data/trimmed_reads/${base}.trimmed.fastq" \ -S "data/alignments/${base}.sam"

and I got these results (it increased almost 1%) reads processed: 85645 reads with at least one alignment: 959 (1.12%) reads that failed to align: 84686 (98.88%) Reported 59442 alignments

should I change the parameters???

Answer 1

[Kevin]

Because the read length goes to 23 bp on average after trimming, I'd guess your library quality is fine and the reads are structured as expected. That leaves the alignment step as the culprit. (Bowtie2 should work fine for this application). I'm guessing something went wrong when you built your index.

[Ant]

Hi, I rebuilt the index and also used the inspect function and apparently everything is correct. I also don't know what could I be doing wrong.

The inspector output:

---

. > cel-let-7-5p MIMAT0000001 Caenorhabditis elegans let-7-5p TGAGGTAGTAGGTTGTATAGTT . >cel-let-7-3p MIMAT0015091 Caenorhabditis elegans let-7-3p CTATGCAATTTTCTACCTTACC

[Kevin]

Just so we can get a feel for what your trimmed fastq looks like: what is the output of this command?

sed -n '2~4p' data/trimmed_reads/${base}.trimmed.fastq | sort | uniq -c | sort -nr | head

This should show your most common read sequences

[Ant]

Undestood. This was the result. Do you believe tha the total alignment I achieved is the total possible result for the sample? The rest could be dimers or something like that?

>44912 TGGTAATACGACGTACTTAGTGT

>6251 TTGAGGATATGAGATCGAGACGACA

>618 CTTTCGCTTCGTCATCTACT

>451 TAGCAGCACGTAAATATTGGCG

>369 GTCGACGCTAGCGGATCCT

>350 CCGAAGAAGTTGAAGATCGTC

>300 GCCTGACCGAACTGTGTCT

>264 ATTCCGGTTGTTCTTCTCT

>231 TTGCATAGCTGGAGAAGTGATC

>231 CCGAAGAAGTTGAAGATCGT

[Kevin]

True dimer would show up as empty sequence after trimming. To me, this looks like something atypical happened during library prep. These aren't known small RNAs. Could be side products formed during RT. Could be a bad library that somebody tried to salvage. I would ask wet lab for final library traces, plus info about how much starting material they used and how many PCR cycles. Also ask if they did anything atypical like PNK treatment or a gel cut or Pippin size selection at the end.

[Ant]

Okay, I will do that. Thanks a lot for your time and help. I will talk to them and get back with the answer.