Options for PGM Variant Analysis

[HeidiJTP]

I am preparing to order barcoded primers for a sequencing project on the PGM, and would appreciate some feedback on the options for data analysis before I make this purchase. We are sequencing 4 loci in approximately 65 individuals (diploid, mammal, WGS available). The amplicons will be ~350 bp. We were told this could be done on a single 314 chip.

We are going through a sequencing center, so I do not have access to the Ion Torrent server. If I understand correctly, I could pay to use the Variant Caller through Amazon though. What are the general costs? Does this work when not using the human genome as a reference?

Does anyone have experience with any free programs for variant analysis of PGM data? How about GATK? I hear so much about the need to stick with platform-specific analyses to deal with their particular types of sequencing errors, and just wanted to know what the general thoughts were on that topic.

[thomasblomquist]

I think Nils Homer develop the TMAP software, which is a variant of his BFAST short read aligner that is geared toward the Ion Torrent sequencing error profile, and post-sequencing data analysis for variant calls.

Regards,

-Tom Blomquist

[MrGuy]

You could probably suggest to the sequencing center to have them install a plug in called PathogenAnalyzer. It's on the community. It looks to me to be designed specifically for viral/bacterial amplicons and viral whole genomes. However, I read the manual where you can put in different reference sequences (unlimited?) that it will map against and pick best fit. It also does variant calling. I assume you could fool it into using small regions of the human genome as applicable. I haven't used it myself, so can't help much more.

[snetmcom]

The center should be able to run the variant caller that's part of the server. There is no need for amazon web services. All the torrent server tools are are open source and posted to github.

[HeidiJTP]

Thank you everyone for your suggestions, and for the tip that everything's available on github. I watched a webinar yesterday for the Ion Reporter Software, in which the instructor stated that only the human genome can be used as a reference. I will be working with non-primate species, so that may rule out using their variant caller.

I looked into TMAP and found that it can be combined with the Ion Variant Hunter while specifying your own reference, so that may be an option. PathogenAnalyzer also sounds like it could be useful, so I will explore these further. Thanks!

[jtackney]

TMAP is a software suite. It utilizes 4 algorithms. By default on the torrent server tmap4 is run, which is a variant of bwa-mem. The others are:
tmap1 - bwa-short
tmap2- bwa long read (bwa-sw)
tmap3 - variant of SSAHA

You can pick which ones you want to run if you compile it yourself OR you can run any combination of the the 4 with tmap mapall command.

That said, I've had a wicked time trying to track down who actually upkeeps tmap at lifetech right now. Where did you hear that Nils Homer developed it? I think it was someone internal at Ion Torrent (life tech).

[thomasblomquist]

Email communication from dr. Homer to myself in Oct. 2012.


"Check out TMAP on github for non SOLID reads. It is my next software. N"


I took it to mean it was his development. This email was in response to an email chain regarding the gearing of BFAST index seeds for the ion torrent error profile. It may not actually be a BFAST code base, but I think it is Nils' work based on my email strings.

Btw. Did I mention how much I love the customizability of BFAST? :-)

-Tom Blomquist

[jtackney]

Interesting.... I've never used it actually. Just bwa or bowtie2. Did they get BFAST to work better with Ion Torrent data?

[thomasblomquist]

Interesting.... I've never used it actually. Just bwa or bowtie2. Did they get BFAST to work better with Ion Torrent data?

Honestly, wouldnt know the difference between tmap and bfast. I ended up sticking with BFAST because of how reproducible my sequencing alignment calls were. I work primarily with my own custom amplicon libraries and parsing my reads into subregions being analyzed by unique/custom index masks geared for each fasta reference database, I can get my error rate for alignments into the 1:100,000 and 10 million reads processed on an i7 intel chip with 32gb ram and 8 cores in 1-2 hours. I'm working toward developing some acute care panels so speed, accuracy and reproducibility are of essence.

-Tom

[nilshomer]

Honestly, wouldnt know the difference between tmap and bfast. I ended up sticking with BFAST because of how reproducible my sequencing alignment calls were. I work primarily with my own custom amplicon libraries and parsing my reads into subregions being analyzed by unique/custom index masks geared for each fasta reference database, I can get my error rate for alignments into the 1:100,000 and 10 million reads processed on an i7 intel chip with 32gb ram and 8 cores in 1-2 hours. I'm working toward developing some acute care panels so speed, accuracy and reproducibility are of essence.

-Tom

BFAST and TMAP are not related, as the latter is derived from BWA and customized for Ion Torrent data, while the former was primarly used for SOLiD data alignment and the early days of Ion Torrent alignments. I was the primary developer of both bfast and the extension of BWA (Heng Li is the author of BWA.

URLs:

http://bfast.sf.net/

http://github.com/iontorrent/tmap

TMAP is a software suite. It utilizes 4 algorithms. By default on the torrent server tmap4 is run, which is a variant of bwa-mem. The others are:
tmap1 - bwa-short
tmap2- bwa long read (bwa-sw)
tmap3 - variant of SSAHA

You can pick which ones you want to run if you compile it yourself OR you can run any combination of the the 4 with tmap mapall command.

That said, I've had a wicked time trying to track down who actually upkeeps tmap at lifetech right now. Where did you hear that Nils Homer developed it? I think it was someone internal at Ion Torrent (life tech).

I did develop it, although I am no longer working at Life Technologies.