bowtie --fr/--rf/--ff reporting

Question

[ThePresident]

Hello,

I am mapping Illumina generated paired-end reads with bowtie. By default, bowtie reports --fr, i.e. paired mates are in Forward/Reverse position (which is normally the case for the vast majority of reads). However, I am also interested in reads that map with --ff configuration (both mates on the same strand). Is it possible for bowtie (or alternatively bwa) to report both configurations --fr and --ff as valid?

When running bowtie with -a (report all alignments) and --fr on a subset of reads I get 28,44% of aligned reads. When specifying -a and both --fr --ff, I get 0%.

Thanks

Answer 1

[Devon Ryan]

You have to chose one of --ff and --fr, specifying both makes no sense. At least with bowtie2, these settings only affect what the aligner considers to be properly paired (bit 2 in the flag when you look at a BAM file). Unless you specify --no-discordant, then discordant alignments will still be returned if they exist.

[ThePresident]

Correct me if I am wrong: bowtie will report all alignments, i.e. --ff --fr, broken pairs etc. (unless --no-discordant is specified) in the output (let's say a SAM file). However, if I visualize the alignment on my reference strain, only reads considered valid (by default proper pairs and --fr) will be displayed.

If that is correct, then how can I extract or filter for these reads based, i.e. how can I extract specifically broken pairs, pairs that mapped to the same strand?

I ask this because I am interested in structural variants that I wish to detect with paired-end discordant reads.

[Devon Ryan]

IGV et al. don't filter alignment, you'll see all of them. You don't need to do anything special to find structural variants (from the alignment side) unless you're trying to resolve exact break-points.

[ThePresident]

Interesting. So just to be sure (and please pardon my ignorance on the subject): If I run an alignment with bowtie with let's say 1,000 reads-a, the resulting SAM file will contain all 1,000 reads regardless if they mapped correctly or not (some of them being flagged as correctly mapped, other as discordant etc.). Or, the SAM file will contain only correctly mapped reads (in respect with parametrs such as --ff; -v; -n; -e; -l etc.)

Thanks for clearing this up for me.

[Devon Ryan]

Who knows what bowtie1 is doing, almost no one uses it anymore. In general, if two reads can be aligned in a concordant manner then all other alignments should be ignored, regardless of whether they're returned. You should not be using bowtie1 with the -a option.

[ThePresident]

Thanks. Actually, I -a just for the purpose of the example (trying to understand how bowtie operates). Is bowtie2 different in that regard?

[Devon Ryan]

bowtie2 doesn't have an -a option. In bowtie2, if a pair can align concordantly then that's what's reported. If it can't then it'll try whatever discordant alignment it can reasonably find. That's generally the most reasonable strategy.

Answer 2

[mastal511]

Illumina paired-end reads are generated with the --fr configuration, one read of the pair is sequenced from the forward strand, and the other read is sequenced from the opposite strand.

[ThePresident]

Correct. But if you align them, paired-end mates can align in --ff configuration if, for example, there is a genomic rearrangement (inversion of a segment) in the sequenced strain compared to the reference.