Anyone has experience with Bowtie2?

[LP_SEP23]

Hi All,
my name is Lucia, and I've been lurking around for a while but this is my first post.
I have a pretty big ChIP-seq experiment to analyze, and initially did the alignments with Bowtie. The stats were not that great, and since I have 100bp reads, I decided to redo the alignments with Bowtie2
I do get a lot more mapped reads, but the quality scores seem a lot lower
The programs compute the scores differently, so I am not sure if the scores are really "lower".
Does Bowtie2 increase mapping at the expense of quality?
Or is the scoring scheme so completely different that the quality scores are not comparable at all?
thanks

[epi]

Sorry can't help you with your specific problem.

Though in my experience, bowtie works much better than bowtie2. Apart from some bugs, many options are missing (strata/unique alignments). I shared my opinion with authors as well.
Must mention that I have 50bp reads and my final goal (ChIP-Seq) may be different from you.

[NicoBxl]

Maybe try bwa

[sdriscoll]

Depends on what score you are talking about. The MAPQ score from bowtie2 has to do with the uniquness of the alignment and not necessarily the quality of the alignment. You want to look at the AS attribute in the extended attributes of the SAM files. Also read up on this section of the bowtie2 manual:

http://bowtie-bio.sourceforge.net/bo...r-more-similar

By default bowtie2 is allowing no mismatches in the seed (22 bases for default end-to-end alignment) and a minimum alignment score of -60.6 (with a max of 0) for 100bp reads. Default penalties are something like -6 for a mismatch at a high quality base an -11 for a length-2 gap. Gaps in reads and reference are scored the same. Specifically you get a -5 for a gap open and -3 for each base. So a length-1 gap is -8, length-2 is -11, etc. So with a min score of about -60 you could have as many as 6 mismatches past the seed or all kinds of combinations of gaps and mismatches. You can control all of those parameters. The min score setting is defined as -0.6 + -0.6*L where L is you read length. If you want to set that value you do so with --min-score L,-0.6,-0.6. So you can figure out what you want your min score to be, like say -40, and work backward from that to figure out what values to put in the setting.

[adaptivegenome]

Mapping quality is valuable for variant calling. If the quality score does not correspond to accuracy of alignment, this is a big problem I think.

[sdriscoll]

Mapping quality is valuable for variant calling. If the quality score does not correspond to accuracy of alignment, this is a big problem I think.

You're right about that. That's probably why it's better to use BWA for those alignments. I don't know of any advantage other than speed for using bowtie2 over BWA. I haven't done any hardcore comparisons between the two. I do like BWA's alignments more though...not sure why.