Inordinate Number of Reads Being Filtered

[Dynamac]

I am losing about half of my data from the TrimBack Valley and Signal Intensity Filters (numTrimmedTooShortQuality). Has anyone had this problem, and if so, what are your solutions and/or recommendations?
Any comments at all would be appreciated, as I am still a newbie with this system, and support from Roche has been spotty at best.

Thanks!

Barry

[vamosia]

Which pipeline are you using? The ends of the reads may be of lower quality, which may caused reads to be filtered out.

If your still interested in the those rejected reads anyway, you could try the standard shotgun pipeline (which supposedly is less stringent compare to the amplicon pipeline)

[Dynamac]

Yes, we tried the shotgun pipeline and that roughly doubled our number of bases called.

Thanks!

[vamosia]

Great, keep in mind that the amplicon pipeline does several additional correction to help reduce homopolymer error (amplicon library are generally very bright, hence more chance of producing homopolymer error).

While the using the shotgun pipeline to process amplicon samples gives more data, it may or may contain more homopolyer error. Just something to watch for.

[ajthomas]

In my experience, the extra data you get by processing amplicon data with the shotgun pipeline doesn't help. The extra data doesn't ususally equate to more information because the lower quality of the data offsets the reduced quantity. As for homopolymers, I do find more errors in homopolymer regions. Of course, your experience may differ from mine. It's probably worth it to try it both ways and see if that extra data helps or not.

There are a few TCBs relating to amplicon sequencing that you should read if you haven't already. There's one (sorry, I don't remember which one off the top of my head) that talks about the what makes the amplicon filtering pipeline different. You should certainly read that if you haven't already.

[pmiguel]

But the amplicon pipeline does not do quality trim-back. It tosses reads that don't meet its quality criteria, but the reads it keeps may still have low quality 3' sequence in them.

--
Phillip

[RCJK]

One of the recommended filtering strategies is to use the amplicon pipeline, but change the setting so that it will trim the reads instead of discarding them. This is in the latest (July 2011) guide to amplicon experimental design.