What Is The Fastest Alignment Tool To Map Reads To A Big Genome Over 4Gb?

Question

[Dejian]

We assembled a genome around 5GB. Now we are trying to map the Illumina reads to the genome. There are many aligners that can do this job. But which is the fastest one? Have anyone benchmarked any of these tools?

The list of aligners is as follows: short-read aligners.

Answer 1

[Istvan Albert]

Asking for the fastest tool is posing the question the wrong way.

What you need is finding a fast enough method for your data. As Jeremy points out there are many questions that need to be answered and you will need to run your own tests to identify which tool is fast enough and provides the necessary coverage as well. Your genome is likely to contain vast streches of low complexity regions - that could make some approaches perform differently.

Start with the known fast tools like bowtie and SOAP and see what these do. Also read the literature: Comparative analysis of algorithms for next-generation sequencing read alignment

Answer 2

[Ido Tamir]

There are some comparisons the latest to my knowledge is Ruffalo. But is being the fastest really an important criteria, if the difference to more accurate results is 2x running time?

[Dejian]

Currently we want to get a rough result of the coverage, so speed is the first consideration. Thanks for the paper recommended.

Answer 3

[Jeremy Leipzig]

you need a lot more criteria to make this a fair fight
here are some just off the top of my head

1. number of mismatches
2. memory ceiling
3. paired-end?
4. quality-aware?
5. sensitivity

[Dejian]

Currently we want to get a rough result of the coverage, so speed is the first consideration. Thanks for your suggestion.

[Istvan Albert]

your comment indicated that you do not understand that accuracy and sensitivity are not independent from coverage. Increasing these will necessarily slow down the method.

Answer 4

[predeus]

I think it would be interesting to bump it up in 2018.

What is the fastest aligners out there? I have heard (I think Heng Li wrote it somewhere) that with changes in computer architecture, hash-based aligners are coming to be more and more interesting, compared to FM-index based. I am about to try our SNAP, which promises up to 20x increase compared to bowtie2 and bwa.

Like Istvan rightly pointed out, aligner has to be fast for your application. But sometimes speed could be a limiting factor - e.g. when you are teaching a class and want to show something pretty quickly, and not wait for 30 minutes.

Answer 5

[Gjain]

I use Novoalign (http://www.novocraft.com/main/page.php?s=novoalign) which is good and fast.

[brentp]

is novoalign really fast? I've found it to do a good job, but not especially fast.

[Gjain]

it depends on your definition of what you call fast. If I run my mapping (HG19) pipeline on our cluster with around 800 nodes ... it normally takes 4-5 hours.

[Rm]

@Gjain: Do you use MPI version? or ...

[Gjain]

just normal version

[liuyanhu005]

Hi，I‘m using a software named LSC,which need novoalign to align.But I can’t download novoalign.And I can't open tihs URL.Can you send it to me? My email address is [e-mail]. Many thanks！

[Sparks]

Hi, My apologies for reviving an old thead but I've receently become aware that the default options in LSC for Novoalign are far from ideal.

If you are using Novoalign with LSC you need to use -r Exhaustive not -r All. This will improve results dramatically and should be better than the other aligners.

For 100bp reads try.. novoalign_options = -c1 -r Ex 1000 -t 120 -F FA -g 0 -x 20 -o sam

If read length is different adjust -t accordingly. On Cerebellum reads used by Kin Fai results are 40% better with 30% reduced run time vs BWA using LSC 0.3.1

Best, Colin

[Gjain]

This is very useful ... thanks Colin

[Gjain]

hi, please use this page to review the licence agreement and you can download the one which suits your needs best. http://www.novocraft.com/main/page.php?id=968

[liuyanhu005]

Thank you a lot!