Bowtie2 Local Alignment Not Working

Question

[lvl2111]

I have an amplicon library, which I am trying to align with bowtie2 but I am having issues.

My reads have been trimmed to removed the 5' and 3' adapters, which ideally should result in 20bp trimmed reads. However, sometimes, there are reads that are 22-23bp. The indexes that I am aligning to are 20bp gRNA barcode sequences, matched to a gRNA barcode name.

When I run bowtie2 with end-to-end, it works quite well and I get ~85-90+% alignment.

bowtie2 --norc -N 1 -x library/crspri_grna-index -U inputfile.fastq.gz -S outputfile.sam  >  logfile.sam.log 2> errorfile.sam.err

Results:

3474015 reads; of these:
  3474015 (100.00%) were unpaired; of these:
    401777 (11.57%) aligned 0 times
    2943712 (84.74%) aligned exactly 1 time
    128526 (3.70%) aligned >1 times
88.43% overall alignment rate

However, some reads are not trimmed to exactly 20bp, and therefore I am getting biases in my results. I therefore need to run it on a local alignment.

However, very oddly, when I run bowtie2 with the --local or --very-sensitive-local parameter, I get 0 aligned reads.

bowtie2 --norc --local -N 1 -x library/crspri_grna-index -U inputfile.fastq.gz -S outputfile.sam  >  logfile.sam.log 2> errorfile.sam.err

Results:

3474015 reads; of these:   3474015 (100.00%) were unpaired; of these:
    3474015 (100.00%) aligned 0 times
    0 (0.00%) aligned exactly 1 time
    0 (0.00%) aligned >1 times
0.00% overall alignment rate

Can someone please help me understand how to fix this to get the local alignment to work? Apologies if I am not posting properly, this is my first post.

If it's useful, here is an example of my fastq file. The majority of the time, my reads are 20bp. However, sometimes, a read that is 22-23bp or longer comes up. These are getting ignored by the end-to-end alignment

AE/EEEEEEEEAA/EEEEEE @SRR12815306.732 732/1 GTACAGGTTCTAACCCGTT
AEEEEEEEEEEE/EEEEEE @SRR12815306.733 733/1 GCTCGAGCTGGAGTTCGACC
/AEEA/A6E/EEEEE/E/EE @SRR12815306.734 734/1 ACACTCATCTCATTTATTCTTGT
AEEEEEEEEEEEEEEEEE6/AEE @SRR12815306.735 735/1 GATGTTTAAATGCTTTTTCG

[GenoMax]

which ideally should result in 20bp trimmed reads.

Please try using bowtie v.1.x. It does ungapped alignments which is what you want with CRISPRseq.

Ideally you may want to use MaGeCK or 2FAST2Q (https://github.com/afombravo/2FAST2Q) meant for CRISPR analysis.

Answer 1

[ATpoint]

These sorts of libraries have a precise length so you don't need to trim adapters by a deficated adapter software. Just hard-trim to 20bp. seqtk can do that. Then run bowtie2 against the barcode library as end-to-end alignment with all gap / penalty parameters set to 10000 to force to only align perfect ungapped matches. Local alignment with wuch short reads it not wise. I used this strategy for these sorts of libraries before and it works well.

[lvl2111]

Thank you both for you responses, this was very helpful.

I compared 2FAST2Q (with no further trimming to 20bp after trimming adapters) versus bowtie2 (bowtie2, after trimming to exactly 20bp as ATpoint suggested).

I found that 2FAST2Q dropped many of my reads (~28% alignment) rate versus bowtie2, which had a better alignment (>85%).

Ultimately, I will move forward with trimming to exactly 20bp and using the end-to-end bowtie2 alignment. Thanks!