454 Titanium 16S HMP primers?

[cliffbeall]

We are trying to get started analyzing 16S RNA using the primers and protocols posted by the Human Microbiome Project (see here : V1-V3 and V3-V5 primer sets). Our problem is that we are getting the majority of our reads removed by the "short quality" filter. We started with 2.5M reads passing the key sequence, but about 1.8M are failing the short quality filter, and we only get 295,000 passing all filters. This is with the 2.5.3 version of the 454 software, which was actually an improvement on the previous version, which only gave ~100,000 passing filters.

On the theory that there might be short fragments in the sample, we repurified everything on Ampure and ran 2/8s of a plate but didn't see any improvement.

Is anyone else using these HMP primers? Have you seen this short quality filter problem? Are there adjustments to the sample prep or the analysis software that we can make? Could it be something with the density of beads, are we loading too many? Any help would be appreciated, thanks.

[nickloman]

You may find my recent blog post on this subject helpful

http://pathogenomics.bham.ac.uk/blog/2011/05/curious-results-from-454-amplicon-processing/

[flxlex]

Indeed, study the blog post, as it looks like you need to recall the bases with changed filter settings...

[cliffbeall]

Nick, thanks for the link. I actually have read your blog and even commented once but hadn't seen that post. I am going to get with our core and have them try those adjustments and will report back.

[cliffbeall]

Well, I tried editing that template file and re-running and even fewer reads pass the filters, though now they are lost in the dot filter rather than the short quality. I am going to try the shotgun and the modified approach here to get all keypass reads today.

[Palecomic]

Have commented on Nick's blog and updated here: http://seqanswers.com/forums/showpos...0&postcount=27 we're using the V3-V5 primers and you can shift some of those medium length reads upwards whilst using the shotgun (or alternative amplicon processing) by changing the emPCR conditions a bit

[cliffbeall]

Thanks for the input, Mike (Palecomic). I can say that I tried all of those filter changes I mentioned earlier and nothing worked, even the modification that was supposed to give all key pass reads doesn't work as advertised. In fact every modification I tried gave fewer passed filter reads, although it varied which filter was catching them.

We looked for "cryptic" small fragments based on information from another sequencing center - using the quality control PCR from the short fragment removal procedure (see Roche Technical Bulletin TCB No. 2011-002), but haven't found short fragments in our libraries.

We are just about to do another run, with the V1-3 primers only, shortening the initial sample PCR from 30 to 25 cycles, with further modifications to the EM-PCR - I think cutting the primer concentration and using fewer molecules per bead. If it works I will find out the exact differences from our core folks and post here.

[Biohazard0806]

Hello cliffbeall,

we also have the problem with the reads removed by the short quality filter by using the GS Junior system. It would be nice to know if and how you solved that issue.

Thanks

[cliffbeall]

Things are working better. Anthony.287, who runs the instrument, posted info on a recent run here. The crucial parameter seems to be copies per bead in the emulsion PCR, reduced from 2 to 0.08 (!).

Our experience is with the Titanium though, maybe someone can comment on junior-specific issues, if any?

[Anthony.287]

Hello cliffbeall,

we also have the problem with the reads removed by the short quality filter by using the GS Junior system. It would be nice to know if and how you solved that issue.

Thanks

I don't think that we ever actually had small fragments in Cliff's libraries; the short quality filter category actually includes a number of filters. So, reads that were failing the valley filter, for example, were included in the short quality filter, leading all of us to think that we had short fragments, even though we couldn't find them. As the enrichment percentages got closer and closer to 8%, the number filtered out with short quality went down. We had mixed beads and overloaded PTPsd, not short fragments. Another tip I discovered recently is to underload the PTP when sequencing amplicons; this can help reduce crosstalk between wells. For example, load 1.7M beads instead of 2M. It does seem to help.