Bowtie2 --score-min option versus default alignment

[Pradhaun]

Hello All,

I am using bowtie2 for mapping illumina reads (100 bp). Since I am trying to align the various different plant lines of same species to the reference genome, I would like to play with the mismatches within each read. Because, I cannot expect the 100 bp of a different plant line to match exactly with the reference genome. Therfore, I was using --score-min option, thereby I can manipulate the alignment score.

For eg, bowtie2 --score-min L,0,-0.3, -x my_ref_index -1 read1.fq -2 read2.fq -S output.bam

The above command allows 5 mismatches (1 mismatch = -6 penalty).

But I am not very sure whether this is the right way to do it. I would really appreciate if anybody can please provide their suggestions. And is the default bowtie2 alignment differs from alignment with --score-min option, other than allowing 5 mismatches?

Thanks!!

[dpryan]

The default is L,-0.6,-0.6, so your option will allow fewer mismatches than the default, which probably isn't what you want. I should also note that the mismatch penalty isn't always 6, it varies with the base quality.

[Pradhaun]

Thanks for the quick reply drpryan.

And thanks for pointing about default score. I will recheck my scoring system.

I should have explained my question in a better way. Actually my question is, if I would like to see the alignment rate in a range of 0 to 5 mismatches, should I have to use --score-min L,0,0 for 0 mismatches or can I use the default alignment without any score-min option?

bowtie2 --score-min L,0,0 -x index -1 read1.fq -2 read2.fq -S output.sam
or
bowtie2 -x index -1 read1.fq -2 read2.fq -S output.sam

I am getting higher alignment rate for the latter compare to the former command (with --score-min option). As per my understanding, the above two commands are not doing the same thing. If yes, can you please explain the reasoning behind it? Why I am getting higher alignment rate when I am not using --score-min? I would really appreciate if you can clarify about this.

Thanks!

[dpryan]

It's simpler to just use the default settings and use grep to see how many alignments have a given edit distance (see the NM tag). This won't be perfect, since an InDel will be treated like a SNP, but it'll be a reasonable enough approach.

Not specifying --score-min is the same as specifying "--score-min L,-0.6,-0.6", so it's unsurprising that you get a higher alignment rate when you allow mismatches than when you don't.

[Pradhaun]

Thanks dpryan.