MiSeq best result?

[pmiguel]

I am calling out all who think they have good MiSeq runs. Gaze at the perfection of this run and weep at the inadequacy of your wretched instruments:



But seriously...
Accidently clustered at a higher density than normal. But the results were still good. Raw cluster density was 1154 Kclusters/mm^2. 8.8 billion bases total. 6.8 billion bases >Q30.



Kind of interesting, this indicates that 0.6x0.4 = 24% of read pairs have zero sequencing errors in them!


And, just to add historical perspective, this is around the amount of data generated by 13 GS-FLX runs. (And we would bill out a GS-FLX run at 5x the cost of a MiSeq run.) Also, at 80 kb per 3730XL run, we would need 13,000 of them to equal the output of this single MiSeq run. That is 3 years of non-stop runs on a 3730XL!

--
Phillip

[GenoMax]

That is .. normal genomic DNA, I assume.

How about running some 16S samples.

[pmiguel]

That is .. normal genomic DNA, I assume.

How about running some 16S samples.

No, an RNA seq library.

We will be doing some 16S fairly soon.

--
Phillip

[yaximik]

I mentioned in other thread that I was getting decent results at density of as much as 1500. My problems is associated more with short libraries made from ancient DNA, so pushing longer than 2x151 cycles makes little gains. But with density of 1100 I got about 75% of >30Q with 2x151 cycles.

[Jean]

How much PhiX spike-in?

[yaximik]

If this is question to me, I do not spike-in as I am using only fastqc framework - genomes are too big to be processed on-board

[westerman]

No phiX on our (Purdue's) end. We ended up with about 0.1% of the reads which potentially map to phiX but then we always seem to get this low-level matching. As for the overall stats:

33,676,040 raw reads
8,452,686,040 raw bases
251 length

33,156,566 reads after adapter and quality trimming
7,559,421,635 bases after trimming
30-251 length with a mean of 227
1% of the sequences lost to trimming
10% of the bases lost to trimming

It is an incredible amount of data, at least for those of us who have been around for a long time. Equivalent to 2x human genome coverage for, what, $3000 or so.



Overall very comparable to other miSeq runs we have done.

[pmiguel]

Rick,
What? We did spike in phiX!
We ostensibly spiked in 1% phiX -- but it turned out that only 0.3% of the reads derived from phiX.
phiX gives the % perfect read plot.

--
Phillip

[GenoMax]

We have not had any problems getting the clusters to register without spiking in phiX with 16S runs. It is the quality values assigned to the bases (in latter half of 2 x 250 bp runs), where the problem occurs. Sequences being generated have been acceptable for the end users with no problems about the actual base calls.

Would be curious to see how you fare with your 16S runs. I assume you will be spiking in phiX.

[westerman]

In reply to Phillip:

Ooops. Don't know what I was thinking. Obviously I wasn't. :-( Should have looked at the lab notebook before spouting off to the world.

[pmiguel]

We have not had any problems getting the clusters to register without spiking in phiX with 16S runs. It is the quality values assigned to the bases (in latter half of 2 x 250 bp runs), where the problem occurs. Sequences being generated have been acceptable for the end users with no problems about the actual base calls.

Would be curious to see how you fare with your 16S runs. I assume you will be spiking in phiX.

We will likely spike in genomics libraries of some sort. That way some useful data is generated from the genomic 'ballast' library.

So you are saying that the quality values assigned are inaccurate (lower than they should be)?

--
Phillip

[bstamps]

We should be getting back our MiSeq 16s results within the next day or two- did a 50% spike-in of gDNA under the recommendation of the center. Right now our SOP is to find some genome in our lab that needs sequencing and use that to up the diversity on the run. Otherwise it's 90 barcoded samples...I'll let you guys know how it went when we get it!

[GenoMax]

We will likely spike in genomics libraries of some sort. That way some useful data is generated from the genomic 'ballast' library.

So you are saying that the quality values assigned are inaccurate (lower than they should be)?

--
Phillip

That is the assumption we are going on. We (meaning the people in the wet lab) are using custom primers, so genomic "ballast" library is not an easy option.

[creeves]

Our record

Was the incredible amount of data you got from one run using an upgraded MiSeq? We have not had the upgrade yet. Our record with RNAseq libraries has been 2.5 Gb of calls >Q30 (1781 K/mm2, 78.5% pass, 13.37 M raw reads, 2x150bp). RNAseq libraries allow higher resolvable cluster density than Nextera yeast genome libraries. Our record with Nextera libraries is 2.4 Gb of calls >Q30 (1262 K/mm2, 90.1% pass, 9.98 M raw reads, 2x150bp).

We suspect that the higher cluster density with RNAseq libs is due to there being fewer longer fragments in the library, which would presumably form larger clusters that would be harder to resolve. We would love to know about tricks to achieve higher cluster densities with Nextera libraries. We already modified the PCR in the protocol by reducing the extension time from 3 min to 1 min and adding two more cycles (7 total).

[pmiguel]

Was the incredible amount of data you got from one run using an upgraded MiSeq?

Yes.

We suspect that the higher cluster density with RNAseq libs is due to there being fewer longer fragments in the library, which would presumably form larger clusters that would be harder to resolve. We would love to know about tricks to achieve higher cluster densities with Nextera libraries. We already modified the PCR in the protocol by reducing the extension time from 3 min to 1 min and adding two more cycles (7 total).

You want to remove the larger fragments, you mean? You could do an Ampure upper cut, right?

--
Phillip