Roche software update 2.0.01.12

[Tuxido]

Did anyone already install the new software update (2.0.01.12) from Roche?

We're unexpectedly seeing a nice increase in mapped reads and coverage. So we're rerunning all previous mappings.

They also changed something in the mutation detection software, but although there are now more mutations detected, we don't see a real accuracy improvement when comparing the sequence data with our SNP array results.

Anyone else tried it yet?

[westerman]

We are running 2.0.01.12. There is really no reason not to since the upgrades from Roche are so easy to install. I have not seen any great benefit to the new version but on the other hand we are (a) having technical problems in the lab and (b) use so many non-model organisms that we are just glad to get any data from in the first place much less the best data.

[kmcarr]

I "attended" the webinar yesterday presented by Jim Knight at Roche on this update and the pending updates (due in the Fall). This update (2.0.01.12) comprises minor algorithmic updates and bug fixes. They did not recommend that labs reprocess their data with this update. The release due in the Fall will contain major algorithmic updates and labs may want to reprocess some runs with the new software.

[flxlex]

I do not know if anybody has noticed an extra peak in their Titanium read length distributions around 400 bases? We do, and according to what I heard from 454, this peak is the result of a bug and this update will remove it. The reads in the ~400 bases peak map poorly, I understand (we use our reads mainly for de novo assembly).

454 recommended not reanalyzing previous runs entirely, but regenerating the sff files form a processed run using the runAnalysisFilter command.

Anybody who has done this and seen an effect on the read length distribution?

[Tuxido]

We did see the additional peek but gathered that is was something that happened in the lab.
I read the release notes of the 2.0.1.12 update but saw no notes that signalProcessing was also effected by this release. We can't wait till Fall anyway, because we're just about to submit our results. We just remapped all of these results with the 2.0.1.12 software to make sure that we didn't miss anything. Luckily mapping is a lot faster with the new version

I'll start a new signalProcessing run and have a look whether anything has been changed and post any interesting results.

[dan]

Does this look to be affected by the above mentioned bug?




i.e. the peak around 410 bp

Given that I only have the SFF's for these files, is there anything I can do to fix this problem?


Cheers,

[flxlex]

Improved read length distribution after reanalysis

Hi,

We have now reanalysed our runs with 2.0.01.12 and see some marked improvements in the read length disributions. Not all runs benefit as much, but in the ones I looked at, the extra peak around 400 bases dissapeared, and many more reads ended up around 500 bases. Some examples:

Nice improvement:


Very nice improvement:


Not so much improved:


Any else seen the same? These are all samples from shotgun libraries from the same source, I need to start checking other kinds of samples...

flxlex

[dan]

Thanks for posting these results flxlex. To be clear, they are obtained by re-running the 2.0.1.12 runAnalysisFilter on the data generated during the image processing step (.cwf files)? Or did you re-generate the .cwf files from the images?

I'm tempted to contact my sequencing centre to ask them to re-analyse the runs before they potentially delete the processed image files (.cwf). If this will give us more longer runs of better quality from the same data, then I think it should be worth doing.

Thanks again,
Dan.

[Tuxido]

I think this is only the signalProcessing (the filtering of the .cwf files).

We tried the same on a few (3 ) runs but did not see any real improvement:
- average read length decreased slightly, median read length slightly increased.
- We still see the extra peak and it is no less than it used to be
- maximum read length is increased to up to 1400bp which is quite a bit suspicious.

Also I've been told by Roche to wait for the Fall software release which will make it more worthwhile do redo the signal processing.

[flxlex]

Thanks for posting these results flxlex. To be clear, they are obtained by re-running the 2.0.1.12 runAnalysisFilter on the data generated during the image processing step (.cwf files)? Or did you re-generate the .cwf files from the images?

Indeed, We ran the v2.0.02.12 runAnalysisFilter command from inside the D_ folder of the original data, thereby using existing .cwf files.

[RajAgainstTheMachine]

Update 2.0.01.12

It looks like I am running into the same problem. Where does one get the update from?

[westerman]

Where does one get the update from?

Roche.

This question has, as I recall, been brought up on this forum before. Roche has the software but seems willing to give it to anyone especially if they go through their sequencing facility.

[maubp]

Roche.

This question has, as I recall, been brought up on this forum before. Roche has the software but seems willing to give it to anyone especially if they go through their sequencing facility.

Indeed, see http://seqanswers.com/forums/showthread.php?t=114

[dan]

I heard that version 2.3 will be out in the middle of next month. If you're planning to do a lot of re-processing, it may be worth hanging on for that version (assuming Roche keep to their rumoured release schedule).

Anyone got details?

[Tuxido]

I got a small overview of what the main new features will be:

- transcriptome assembly and mapping modes for the GS de novo Assembler and the GS Reference Mapper,
- parallelization of assembly, mapping and amplicon variant analysis (means: run the software with more than one CPU)
- structural variation reporting for the GS Reference Mapper
- Titanium amplicon variant analysis (400 bp Amplicons)