Lots of chimeric transcripts Ion PGM®

[bernardo_bello]

Hello Bioinformatic community,

We have recently sequenced a bacterial transcriptome with 316 chip from IonTorrent (1.5 million sequences). After filtering low quality data and trimming adapters we noticed that only 51.33% sequences were mapped on reference genome. Looking for the unmapped sequences we can see that most of them are chimeric transcripts, so impossible mapping for them an also causing bias on results. Also many of the unmapped are sequences lacking homology in 20% of the starting sequence.

I would like to know your opinion about it.
Should I have to move to 454 or Illumina? Our Sequencing Department have no idea of why we have so many chimeras.

Thank you, Bernardo

[sulfobus]

Is the read length disitribution for the chimeras smooth, or does it have many spikes?

[bernardo_bello]

Hi sulfobus and sorry for the dealay in response,

The real distribution could be smother (See the attached picture).

The problem is with other analysis in fastQC (See link for complete FastQC file of unmapped sequences).

https://docs.google.com/folder/d/0B8-ZAuZe8jldaUNFSmEzTGVPZFU/edit

Also, I have overepresented sequences.

Bernardo

[sulfobus]

We've had chimeric sequences but they were in distinct groups with the same lengths. From your distribution it looks like most are spread out, although there are two peaks. In our case it was a PCR artefact, stemming from sequence similarity allowing chimeras to form during the plateau phase of the PCR (we used fusionprimers to attach adapters). We solved it by running fewer cycles.

Our application appear to be quite different from yours though, so I don't know if that's of any help.

[bernardo_bello]

Thank you sulfobus. I will tell you how this story ends. Ion Torrent technical support is revising my data.

Bernardo

[vbiaudet]

chimeric reads PGM

I have exactly the same problem with Ion Torrent reads for RNASeq and many chimeric reads after Bionformatic analyses with Bowtie2. Did you obtain an answer from Ion Torrent? Because with Illumina this problem was not here?

Thank you, Veronique

Hello Bioinformatic community,

We have recently sequenced a bacterial transcriptome with 316 chip from IonTorrent (1.5 million sequences). After filtering low quality data and trimming adapters we noticed that only 51.33% sequences were mapped on reference genome. Looking for the unmapped sequences we can see that most of them are chimeric transcripts, so impossible mapping for them an also causing bias on results. Also many of the unmapped are sequences lacking homology in 20% of the starting sequence.

I would like to know your opinion about it.
Should I have to move to 454 or Illumina? Our Sequencing Department have no idea of why we have so many chimeras.

Thank you, Bernardo

[bernardo_bello]

Hi Veronique,

When I asked them for this problem to LifeTech NGS Support http://ioncommunity.lifetechnologies.com/welcome said me that fusions transcripts can be caused by a bad library preparation.

The also said that BWA is not the recomended mapper to Ion Torrent. So I tried the recomended one: TMAP https://github.com/iontorrent/TMAP or you can find it here https://test.g2.bx.psu.edu/ and when I repeated the mapping step, I got magical result, from 30% of mapped reads to 99%. At this point I am happy to have this result but a bit worried about the possible bias caused by the artificial mapping: as I read in https://github.com/iontorrent/TMAP TMAP splits fusion reads and maps the splited reads, but only the longest part of the read, discarding the shortest, if I understood well.


We are now sequencing more samples. If I have the same strange result I will write LifeTech NGS Support to ask if this fusion transcripts are causing bias in my results.

Regards, Bernardo

[Eric@Kapa]

Depending on your lib prep method, the % chimeric lib fragments will be very dependent on adapter:insert ratio. Because insert molecules are 5'-phosphorylated, they can be ligated to one another to form concatemers, which is what would happen in the absence of adapters. It is only the relatively high concentration (ratio) of adapters versus inserts that prevents multiple insert molecules from being ligated to one another.

In our experience, total molar conc of adapter (A and P1) is optimal at ~10:1. Much lower than 5:1 and you start to see an increase in insert fusions (chimeras). Too high, and you lose ligation efficiency, and struggle to remove all unligated adapter in the post-ligation cleanups.

[vbiaudet]

chimeric data

Hi Bernardo,

The use of Tmap is not always a solution... if you need assembly data you can't with chimeric reads... the difference of quality of reads with Hiseq data is very important.

I also found that 20% of reads map rRNA even with a protocol specific to mRNA (polyA detect), did you detect the same problem?

bye, Veronique

Hi Veronique,

When I asked them for this problem to LifeTech NGS Support http://ioncommunity.lifetechnologies.com/welcome said me that fusions transcripts can be caused by a bad library preparation.

The also said that BWA is not the recomended mapper to Ion Torrent. So I tried the recomended one: TMAP https://github.com/iontorrent/TMAP or you can find it here https://test.g2.bx.psu.edu/ and when I repeated the mapping step, I got magical result, from 30% of mapped reads to 99%. At this point I am happy to have this result but a bit worried about the possible bias caused by the artificial mapping: as I read in https://github.com/iontorrent/TMAP TMAP splits fusion reads and maps the splited reads, but only the longest part of the read, discarding the shortest, if I understood well.


We are now sequencing more samples. If I have the same strange result I will write LifeTech NGS Support to ask if this fusion transcripts are causing bias in my results.

Regards, Bernardo