Tweet
Hi all!
This is my first post in this community. I'm grateful for any help, and hope to be able to contribute as well (although I'm a bit new myself!).
The short version of my question is this: Do any of you have any idea why merging alignments to a reference split into two parts would give me a different result than an alignment to the whole reference? I set the Bowtie --seed option to be the same, so there shouldn't be anything there due to Bowtie's random selection of multiply aligned reads to report.
For more backstory and background info: I am using Bowtie2 as an aligner to perform pathogen discovery on my paired-end RNA-seq data, in a pipeline not unlike PathSeq (http://www.broadinstitute.org/software/pathseq/). Essentially, I perform an alignment to a human reference, and take whatever doesn't map and try aligning that against the references of several pathogen organisms.
This second step is what's troubling me a bit. I have a FASTA file that contains the references for several organisms. It's too large to create a Bowtie2 index for alone. So, I've split this FASTA file into many parts. As a sanity check, I wanted to see that the result I get from aligning to each part separately was similar to what I would get from aligning to the un-split reference. To do this, I took a small FASTA file and split that into two, and performed three alignments: one to each part, and then one to the whole reference.
When I merged the results of the alignments to each part, I found that I wasn't getting the same alignments. It's not consistent, either, in how they differ: in the whole alignment, I might see 200 reads mapped with MAPQ > 40 to one organism, but in the merged version I'll only see 4. For another organism, it'll be the other way around---I might see 100 reads mapped with MAPQ > 40 in the whole alignment, and 200 in the merged version.
I made sure that in each alignment, I set the Bowtie --seed option to be the same, so there shouldn't be anything there due to Bowtie's random selection of reads to report.
Do any of you have any idea why alignment to the split references gives me different results than to the whole? Happy to provide more information as well, if that's helpful.
Thanks!
This is my first post in this community. I'm grateful for any help, and hope to be able to contribute as well (although I'm a bit new myself!).
The short version of my question is this: Do any of you have any idea why merging alignments to a reference split into two parts would give me a different result than an alignment to the whole reference? I set the Bowtie --seed option to be the same, so there shouldn't be anything there due to Bowtie's random selection of multiply aligned reads to report.
For more backstory and background info: I am using Bowtie2 as an aligner to perform pathogen discovery on my paired-end RNA-seq data, in a pipeline not unlike PathSeq (http://www.broadinstitute.org/software/pathseq/). Essentially, I perform an alignment to a human reference, and take whatever doesn't map and try aligning that against the references of several pathogen organisms.
This second step is what's troubling me a bit. I have a FASTA file that contains the references for several organisms. It's too large to create a Bowtie2 index for alone. So, I've split this FASTA file into many parts. As a sanity check, I wanted to see that the result I get from aligning to each part separately was similar to what I would get from aligning to the un-split reference. To do this, I took a small FASTA file and split that into two, and performed three alignments: one to each part, and then one to the whole reference.
When I merged the results of the alignments to each part, I found that I wasn't getting the same alignments. It's not consistent, either, in how they differ: in the whole alignment, I might see 200 reads mapped with MAPQ > 40 to one organism, but in the merged version I'll only see 4. For another organism, it'll be the other way around---I might see 100 reads mapped with MAPQ > 40 in the whole alignment, and 200 in the merged version.
I made sure that in each alignment, I set the Bowtie --seed option to be the same, so there shouldn't be anything there due to Bowtie's random selection of reads to report.
Do any of you have any idea why alignment to the split references gives me different results than to the whole? Happy to provide more information as well, if that's helpful.
Thanks!