Bowtie : More reference sequences Less aligned reads

[dl432dl]

Hi All,

I use Bowtie1 (version 1.0.0 for MacOSX)

In order to discard some reads, I mapped reads to multiple reference sequences which I want to remove.

I have a problem that Bowtie gave me fewer aligned reads, when I use more reference sequences.

To be specific....
Total sequences I want to discard are 21 sequences, and there are three different groups of sequences, and each groups have 7 sequences.

Group A: A1,A2,A3,A4,A5,A6,A7. -> similarity:53%~99%, seq length: 1550nt
Group B: B1,B2,B3,B4,B5,B6,B7. -> similarity:49%~99%, seq length: 2900nt
Group C: C1,C2,C3,C4,C5,C6,C7. -> similarity:51%~99%, seq length: 120nt
====> Major targets are A1 and B1

By using major two sequences, A1 & B1, I built a index file, and then did bowtie1.
Its log file reports that:
10.00% reads were reported as aligned reads,
00.01% reads were reported as suppressed reads, and
89.99% reads were reported as failed reads.

After that, I did the same process with all 21 sequences : built a index, ran bowtie1.
And I expected that this result would have more aligned reads than former result. However, it was absolutely wrong!

Latter log file reports that:
00.20% reads were reported as aligned reads,
11.00% reads were reported as suppressed reads, and
88.80% reads were reported as failed reads.

I can not understand the reason why more reference sequences have fewer aligned reads.
At least, it should have more or even reads than former result.
Thankfully, # failed reads to align are similar each other.

I used some options :
bowtie `INDEX` -5 1 -n 0 -n 0 -k 1 -m 1 -l 20 --best --phred33-quals --un `UNMAPPED` -q `INPUT` -S `OUT` 2>> `LOG` -t

Thank you!

Jiyoung

[SNPsaurus]

I'm not an expert at Bowtie, but a couple things stand out to me. First, you have -n 0 -n 0 (-n 0 repeated) so is there an option missing and you wrote -n 0 instead?

But the main issue is tied to the -m 1 option. You are telling Bowtie to only report reads that have a single valid alignment, otherwise suppress them. So when you include all the sequences in the index, in which sequences within the group have high similarity, you are making it very likely that Bowtie will find more than 1 valid alignment and suppress the reporting.

[dl432dl]

Make sense! But why so many reads were suppressed?

I'm not an expert at Bowtie, but a couple things stand out to me. First, you have -n 0 -n 0 (-n 0 repeated) so is there an option missing and you wrote -n 0 instead?

But the main issue is tied to the -m 1 option. You are telling Bowtie to only report reads that have a single valid alignment, otherwise suppress them. So when you include all the sequences in the index, in which sequences within the group have high similarity, you are making it very likely that Bowtie will find more than 1 valid alignment and suppress the reporting.

SNPsaurus, thanks!

Yes, your explanation makes sense. So latter index with more reference sequences showed a few reduced failed reads.

BUt still, it is unclear that why so many reads were suppressed ?
Okay, it will be helpful to compare two output files! Thank you!

Jiyoung

[SNPsaurus]

No, the suppressed reads are the ones that are not reported because of your -m 1 option. In your first try (using A1 and B1) very few are suppressed because very few reads align to both A1 and B1. In the second try many more are suppressed because nearly every read that aligns, aligns to A1 and A2 and A3,4,5,6,7, or B1 and B2 and B3,4,5,6,7. When the read aligns to multiple index sequences, then it fails the -m 1 option and becomes suppressed.