Bowtie output format

[kevlim83]

Hi,

I am currently working on aligning paired-end short reads to a reference transcriptome. I am quite puzzled by the output which I am receiving from bowtie.

Currently the bowtie manual specifies that the 7th column of the output gives us the "number of other possible alignments that can be found on the reference as it was with the current alignment".

The following is a sample output from bowtie, i used the command -a to output all possible alignments (i am displaying only columns 1,3,7):

HWUSI-EAS645:1:1:4:2033#0/2 | NM_001317 | 5
HWUSI-EAS645:1:1:4:2033#0/1 | NM_001317 | 5
HWUSI-EAS645:1:1:4:2033#0/2 | NM_022640 | 5
HWUSI-EAS645:1:1:4:2033#0/1 | NM_022640 | 5
HWUSI-EAS645:1:1:4:2033#0/2 | NM_020991 | 5
HWUSI-EAS645:1:1:4:2033#0/1 | NM_020991 | 5
HWUSI-EAS645:1:1:4:2033#0/2 | NM_022644 | 5
HWUSI-EAS645:1:1:4:2033#0/1 | NM_022644 | 5
HWUSI-EAS645:1:1:4:2033#0/2 | NM_022579 | 1
HWUSI-EAS645:1:1:4:2033#0/1 | NM_022579 | 1

Please note this is paired-end alignment, so every two row represents the same alignment.

Look at the 7th column (3rd column in this case), the first line says we have 5 other similar alignments but there are only 4 other alignments from the same read and the last read says there is one other alignment, but where? Shouldn't bowtie print all other possible alignments with the -a command?

Does anyone know how to interpret this data?

Please help.

Thank you,
Kevin

[Ben Langmead]

Currently the bowtie manual specifies that the 7th column of the output gives us the "number of other possible alignments that can be found on the reference as it was with the current alignment".

Hi Kevin,

This is my fault for not updating the 7th-column description with the paired-end case - sorry.

For paired-end alignments, one mate is the "anchor" and the other is discovered by scanning within the window dictated by -I and -X. For the *anchor* mate, the 7th column is exactly as it would be for the unpaired case - it's the # of other possible alignments for that mate having the same set of edits. For the other mate, the number reported is equal to the number for the anchor mate.

Thanks,
Ben

[kevlim83]

Hi Kevin,

This is my fault for not updating the 7th-column description with the paired-end case - sorry.

For paired-end alignments, one mate is the "anchor" and the other is discovered by scanning within the window dictated by -I and -X. For the *anchor* mate, the 7th column is exactly as it would be for the unpaired case - it's the # of other possible alignments for that mate having the same set of edits. For the other mate, the number reported is equal to the number for the anchor mate.

Thanks,
Ben

Hi Ben,

Thanks for the prompt reply.

According to what you say, we should look at the output by considering the "anchor". So column 7 tells us how many other alignments with the same edits. I can understand this fairly well.

However, I think I need to rephrase my question to clarify my doubts. Please refer to the sample output in my previous post. Let us assume rows 1,3,5,7,9 represent the "anchor" pair. Then the 7th col says for rows 1,3,5 and 7, they each have 5 other possible alignments. And using the -a option, I hope to recover all of these alignments. I am assuming that rows 3,5 and 7 are the "other possible alignments" for row 1, since they have similar mismatches. But lets count them, "rows 3,5,7" only make up 3 other alignment, which makes me wonder where did the other two go? Similarly, for row 9, col 7 tells us there is one other alignment but it is not reported. Shouldn't all possible alignments be reported using the -a option? But it seems like it is not happening here. Have I specified the parameters wrongly?

Please help.

Thanks,
Kevin

[Ben Langmead]

Hi Kevin,

Not all anchors lead to alignments, so the behavior you see is expected.

Hope that helps,
Ben

[kevlim83]

Hi Kevin,

Not all anchors lead to alignments, so the behavior you see is expected.

Hope that helps,
Ben

Hi Ben,

Oh, yes. I didnt think of that. Yes, it answered my question.

Thanks very much.

Kevin

[kevlim83]

Just one more question though, is it possible to force bowtie to ignore the counts which cannot be align in the "mate" pair?

For example, in our example, rows 1-7 will display 3 in col7 and rows 7-9 will display 0 in col7.

Regards,
Kevin