Forum:BWA vs Bowtie 2 (Poll)

Question

[[e-mail]]

Is anyone still using BWA (Burrows-Wheeler Aligner) or have almost all switched to Bowtie 2?

Our bioinformatics team has been evaluating both. BWA is quite slow (and resource hungry) compared to Bowtie 2 although there is a rogue contingent amongst us that are still BWA supporters.

We decided to turn to you to see what the community thinks. So, are you Team BWA or Team Bowtie 2?

Appreciate your input!

Irene
Sequencing.com

[User 59]

This looks like a thinly veiled attempt to drive traffic to your website, and I'm going to close this question as it has no real relevance as a bioinformatics question. No one arriving at this question will find any useful usage statistics.

This might be more tolerably posted in the Forum.

[[e-mail]]

Thank you for your message. I confirm that this post is a real question.

The question in the post is a very real question that our bioinformatics team put forth and the thread has already received several replies over the weekend. As our development team is evaluating BWA and Bowtie 2 and since previous questions/posts relating to this topic are quite old, we were hoping that this post would provide up-to-date insight from the Biostars community.

Thank you.

[User 59]

Regardless of whether this was a post with good intentions, it is NOT a good question for the purposes of this site. Asking a group of people which aligner they favour in a hands up poll is not the kind of question this site was set up to answer. The metrics you will get here are not going to be meaningful (why not ask how many downloads there have been of each package over the last quarter?) and the question is naïve, because it does not address the fact that aligners may be used in preference to one another dependent on downstream pipelines, or sequencing technology used. Someone else has moved it to the forum and reopened it, I will let that decision stand.

[Ashutosh Pandey]

Here we go again :-)

[Ashutosh Pandey]

I have reopened the post after reading Brandon's comments. I have changed the category from "Question" to "Forum".

[[e-mail]]

Thank you, we appreciate your assistance in helping resolve this issue and we'll be cognizant of it going forward.

[Sean Davis]

Bioinformatics is a data science. As such, polls are not really the best way to evaluate the scientific merit of algorithms. We all fall prey to the "everyone is using it" mentality every now-and-then, but there is ample literature on the topic that deserves review. I would suggest a critical review of the literature and available online benchmarks and then asking a more specific question. As an example, you seem interested in the speed of algorithms. See here: http://arxiv.org/abs/1111.5572

[[e-mail]]

Oh no! It sounds like there has been debate amongst the Biostars community regarding this, too. I did check and saw its been a while since this topic was discussed so perhaps this post will lead to some fresh insights!

[Ashutosh Pandey]

You are totally fine. It is always good to get some fresh insights. BTW, BWA-mem (http://arxiv.org/abs/1303.3997) is relatively faster than Bowtie2 in aligning longer reads (> 500 bp). For short reads, Bowtie2 is faster. I usually use BWA for the alignment but I have nothing against Bowtie2. I have extensively used Tophat2 for RNA-seq alignment. Personally speaking I have never found read alignment as a bottle neck in practical settings. In future when getting a genome sequenced will be as common as getting a X-ray, the speed of the aligner will do matter.

Answer 1

[Istvan Albert]

I have had very good experiences with BWA-mem and frankly I would be surprised to hear that it were noticeably slower than Bowtie 2 - although I have not measured it myself.

In my work I have noticed that BWA-mem, when used with default alignment parameters, performs a far more sensitive search than Bowtie 2 - hence it might run slower especially when benchmarked with data of nearly identical reads where the sensitivity of the search is less relevant. One can tune Bowtie 2 to produce the results with the same sensitivity but then the invocation looks like this and runtimes are substantially slower:

bowtie2 -D 20 -R 3 -N 1 -L 20 -i S,1,0.50

just to be able to produce the same quality alignments as BWA-mem. For this reason I pretty much prefer to use BWA-mem.

Answer 2

[JohnBlue81]

Perhaps this post might also give you some more input: Benchmark - Comparison of Different NGS Mappers

[Istvan Albert]

An recent comment is relevant Segemehl: A Fast One-Stop-Shop Mapping Tool