SEQanswers

Go Back   SEQanswers > Bioinformatics > Bioinformatics



Similar Threads
Thread Thread Starter Forum Replies Last Post
TopHat local alignment Giffredo Bioinformatics 8 02-28-2014 12:43 AM
Bowtie2 limit local alignment to 5' end gwl186 Bioinformatics 4 02-18-2014 05:05 AM
Local alignment of short reads: Mosaik, Bowtie2, BWA? carmeyeii Bioinformatics 1 04-19-2013 04:39 PM
Bowtie2 Local Alignment Mode SHeaph Bioinformatics 4 04-18-2013 03:45 PM
gap alignment and local alignment? mingkunli Illumina/Solexa 3 02-19-2009 12:13 PM

Reply
 
Thread Tools
Old 07-06-2015, 06:03 AM   #1
frymor
Senior Member
 
Location: Germany

Join Date: May 2010
Posts: 148
Question bowtie2 local alignment

Hi,

What are the disadvantage of local alignment mode in bowtie2?
We have a data set with very high quality reads. But when running tophat2 with the defaults parameters, even after clipping and trimming for adapter residues, I still can map only ~60-65% of the whole data set.
Than I have tried bowtie2 with the local-very-sensitive parameter and got over 99% of the reads mapped.

I was wondering, if there is anything against using bowtie2 in this modeand why other people are not doing it more often, if the results are so good.
Is it just the longer running time of the mapping step which needed to be addressed here or are there any more factors one must take under consideration, when running bowtie2 in local mode?

I know there are other mappers such as BBMap and subread, that are also able to do a local alignments and works also with shorter reads.

But what are the reason people don't really do local alignments when working with RNA-Seq data?

thanks in advance
Assa
frymor is offline   Reply With Quote
Old 07-06-2015, 06:27 AM   #2
dpryan
Devon Ryan
 
Location: Freiburg, Germany

Join Date: Jul 2011
Posts: 3,479
Default

The only real downside is that it's a bit slower. It's good that you're using the --very-sensitive-local setting, since the defaults for --local are often not that great. For what it's worth, I generally get better results with STAR, though I expect hisat with --very-sensitive-local might work equally well (I've not compared them).
dpryan is offline   Reply With Quote
Old 07-06-2015, 08:38 AM   #3
frymor
Senior Member
 
Location: Germany

Join Date: May 2010
Posts: 148
Default

Quote:
Originally Posted by dpryan View Post
The only real downside is that it's a bit slower.
Thanks Devon for the fast reply. Does it means, I don't have to worry about False positive mapped reads? I mean not (a lot) more, than running bowtie2 in the end-to-end mode or even tophat2?

Is there any reason to worry about downstream analyses, if differential exon usage is involved?
We are also interested in testing the results with DEXSeq. There is a paper from 2011 which sort of comapred different mappers. It shows that bowtie creates a lot of FP junction site predictions.

Quote:
Originally Posted by dpryan View Post
For what it's worth, I generally get better results with STAR...
I have not tried STAR yet, but I would like to try it as well as subread and BBmap, which AFAIK can both works with gapped alignments.

Quote:
Originally Posted by dpryan View Post
though I expect hisat with --very-sensitive-local might work equally well (I've not compared them).
This is also a good suggestion I would like test and compare them.
frymor is offline   Reply With Quote
Old 07-06-2015, 08:47 AM   #4
dpryan
Devon Ryan
 
Location: Freiburg, Germany

Join Date: Jul 2011
Posts: 3,479
Default

I wouldn't recommend using bowtie2 by itself, but with tophat2, since bowtie2 by itself doesn't handle splicing. Regarding junction accuracy, this is probably a better paper to go by: http://www.nature.com/nmeth/journal/...meth.2722.html
dpryan is offline   Reply With Quote
Old 07-06-2015, 08:54 AM   #5
frymor
Senior Member
 
Location: Germany

Join Date: May 2010
Posts: 148
Default

Yes, *that was the problem.
I have tried tophat2 and got 65% mapped reads. This is why i tried bowtie2 with local alignment.
As far as I know, *tophat2 doesn't do local alignments
frymor is offline   Reply With Quote
Old 07-06-2015, 08:58 AM   #6
dpryan
Devon Ryan
 
Location: Freiburg, Germany

Join Date: Jul 2011
Posts: 3,479
Default

Yeah, I don't think there's a good way to use bowtie2 with local alignment on RNAseq data...at least without modifying tophat2. I'd personally use a different aligner for that very reason (well, plus both tophat2 and bowtie2 are a bit slow). You might check and see why end-to-end alignment is proving to be such a problem. If this is something that trimming can remedy (e.g., adapter contamination) then that might be the simplest fix.
dpryan is offline   Reply With Quote
Old 07-06-2015, 10:39 AM   #7
Brian Bushnell
Super Moderator
 
Location: Walnut Creek, CA

Join Date: Jan 2014
Posts: 2,707
Default

The main issue with local alignments in RNA-seq is that, depending on the scoring function, you will lose a lot of the splices. For example, say you have a 100bp read that aligns with a cigar like this:

50=10000D50=

Local alignment shouldn't change anything.

But if you have this:

85=10000D15=

...with the splice site very close to the end of the read, a local alignment would yield "85=" if the score penalty from a 10k deletion/intron was greater than the score of 15 matches. So, generally, I don't recommend local in RNA-seq alignment to a genome - or if you use local alignments, at least check to see the fraction of splice sites captured in local versus global mode. Different aligners have different weighting schemes, and some might not penalize introns at all - particularly, introns won't get penalized by any aligner if you are mapping to a transcriptome since they will not be present.
Brian Bushnell is offline   Reply With Quote
Reply

Tags
bbmap, bowtie2, local alignment, mapping quality, subread

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off




All times are GMT -8. The time now is 11:24 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2018, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO