Re-processing cwf files on GS-FLX

[richardbadge]

We have done a pair of runs on a GS-FLX with Titatnium chemistry... The first time out there was a fluidics failure, and the second time out we have lots of very short reads (50-80bp), despite the facts that we know all our amplicons are >200bp

We'd like to re-process the data from the run to see if aggressive filtering is the problem...

The first hurdle is that it is not clear to me whether the cwf file output from image processing part of gsRunProcessor is overwritten by the signal processing step?

If this is the case I assume we have to repeat the image processing step to get the preliminary cwf files so that we can re-run the signal processing with different parameters...

Alternatively is the preliminary data AND the processed data preserved in the cwf file?

Any help, much appreciated,

Richard

[RCJK]

The way I understand this to work is that the cwf files from the image processing step are what you would use for any subsequent signal processing. The signal processing will output its own cwf files, which will be different than those of the image processing. They are larger and presumably contain more info, but I've never been really sure what their purpose is. The signal processing and image processing results are in different folders and so you won't lose the original image processing data if that is what you are concerned with.

If you are just want to tinker with the filter parameters for the amplicon signal processing, you do not need to completely redo the signal processing. You can just redo the filter steps. Details for this are described in Application Brief APP No. 001-2010 "Amplicon (PCR Product) Sequencing Tips for GS FLX Titanium Reagents" and also in the GS FLX System Software Manual Part B (Section 1.4.3). I've found that this needs to be done with most amplicon runs as the filters are too stringent and like to throw out too many reads. I usually will decrease the vfTrimBackScaleFactor and find this substantially increases the number of pass filter wells.

[richardbadge]

Many thanks!

Hi RCJK,

Many thanks for the prompt response... we are looking at the refiltering the data now.

Also there was suggestion from Roche that we also try the "library" processing pipeline to reduce the aggressive filtering - have you tried this for amplicon libraries...?

Our application is somewhere in between, in that the reagent is a library of amplicons, but the sizes vary quite considerably (from 200bp - 800bp).

Best Wishes,

Richard

[flxlex]

a library of amplicons, but the sizes vary quite considerably (from 200bp - 800bp).

Oh dear... Mixing amplicons of widely different lengths is against the 454 recoomendations and will cause you problems whatever filter settings you choose...

From a technical Bulletin on this:

Since there is a correlation between the number of amplified targets on the DNA library bead and signal intensity, a wider signal distribution will be seen when these mixed libraries are sequenced, potentially lowering the run yield. And when amplicon libraries containing two distinct length populations are combined in a single emulsion, a sub-population of the emulsion droplets will contain two fragments of different sizes. In this situation, the shorter fragment population will amplify preferentially over the longer fragment population during amplification, resulting in either an increase in mixed reads that will be filtered (discarded), or an over representation of the shorter fragments in the sequencing results – and in both cases, resulting in a potential reduction in sequencing yield.

[TonyBrooks]

I've found that this needs to be done with most amplicon runs as the filters are too stringent and like to throw out too many reads. I usually will decrease the vfTrimBackScaleFactor and find this substantially increases the number of pass filter wells.

Hi RCJK
Could I ask what you set your vfTrimBackScaleFactor to? We've run a few amplicon runs with low numbers of reads and we wondered what the consequence of reducing the agression of the filter would be. Obviously we don't want to go too far and get bad quality reads allowed through.
Thanks
Tony

[RCJK]

Hi Tony,

I usually try several different values, decreasing it by 0.5 till I get a high enough number of reads that I am satisfied, while keeping the average quality value still relatively high. I find that a value of 0.5 is what I usually end up going with.

I have tried the shotgun signal processing for amplicons. It'll usually result in the highest number of pass filter reads, but quality and read length will vary. I believe that processing will trim back reads till whatever qualiter filter is met, whereas the amplicon processing will discard the whole read if the filter is not met. Which processing you choose would ultimately depends on the sample, project, what type of downstream analysis is to be done, etc.