Results from 2x250 v2 hardward MiSeq run.

pmiguel · 11-19-2012, 12:32 PM

Finally got the hardware upgrade. Our first run completed over the weekend with a cluster density of just over 800 Kclusters/mm^2. It was genomic DNA libs, so it was an easy test for our instrument.

Here are some pngs from SAV:

I presume the error rate and % perfect reads were derived from the ~1% phiX that I spiked in.

--
Phillip

yaximik · 11-20-2012, 06:18 AM

Could you post also BioAnalyzer electrophoregram, the intensity curves (all Ns, or only C) over the entire run and thumbnails for C at cycles 30, 160, and 240?

NextGenSeq · 11-20-2012, 06:59 AM

This is similar to what we see.

pmiguel · 11-20-2012, 07:35 AM

Quote:

Originally Posted by yaximik

Could you post also BioAnalyzer electrophoregram, the intensity curves (all Ns, or only C) over the entire run and thumbnails for C at cycles 30, 160, and 240?

This is a HS Chip with an overlay of the pre-amp (red) and post-amp (8 cycles) in blue. This was one of 4 libraries on the instrument. It is a little goofy in that we started with only about 10 ng of DNA. So we did the normal DNA TruSeq kit, but used adapters from the RNAseq library so would not end up with bad adapter dimer contamination. Also we sonicated shooting for ~700 bp inserts, then did more stringent Ampures than normal (as low as 0.6X) to try to push the size distributions upwards. Also, I did the denaturation at 10 nM, instead of 2 nM to keep the final NaOH concentration lower.

The other 3 libraries were similar.

Weird new v2 intensity profiles.

--
Phillip

yaximik · 11-20-2012, 08:37 AM

Thanks. This basically confirms what I suspected from my shorter runs. The usable length of fragment is between 1/3 and 1/2 of total length as descending polymerase molecules begin to crowd and interfere with each other in clusters. Or substrate diffusion becomes a limiting factor, or may be both. On shorter fragments this is more visible as cluster numbers begin to drop drastically and intensity curves fall off the cliff. Drop in intensity eventually generates error and aborts the run. In your example bell curves already began to form flattening over their maxima, which are very broad likely due to larger library size heterogeneity. So to take full advantage of 2x250 kit one needs isert size no less than 700 nt, better longer. Perhaps optimal cluster densities with such insert lengths need to be smaller.

pmiguel · 11-20-2012, 08:55 AM

There are lots of possible factors that contribute to worsening results as the number of cycles increases. Polymerases interfering with each other as they approach the flow cell surface seems a little far fetched.

Also, generally one gets somewhat higher quality with shorter amplicons on Illumina instruments. So that doesn't really fit your hypothesis.

--
Phillip

yaximik · 11-20-2012, 11:29 AM

Certainly. The hypothesis is what it is - a hypothesis, more data is needed to prove or disprove, but it seems explaining better what I was observing in my runs than ILM techsupport did. I want to prepare longer libraries with narrower fragment distributions and test the hypothesis more.

HESmith · 11-20-2012, 02:35 PM

I'm in agreement with Phillip - after all, adapter dimers yield great quality scores...

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Demultiplexing MiSeq run - problems	kga1978	Bioinformatics	4	07-21-2016 11:48 AM
MiSeq stopped run	yaximik	Illumina/Solexa	20	03-18-2013 08:10 AM
Need help debugging a faulty MiSeq run	simon_seq	Illumina/Solexa	6	08-15-2012 05:18 AM
Fastq results from MiSeq - which version?	Lilach	Bioinformatics	4	06-19-2012 09:45 AM
Worst MiSeq Run ever?	pmiguel	Illumina/Solexa	0	04-30-2012 05:54 AM

11-19-2012, 12:32 PM	#1
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Results from 2x250 v2 hardward MiSeq run. Finally got the hardware upgrade. Our first run completed over the weekend with a cluster density of just over 800 Kclusters/mm^2. It was genomic DNA libs, so it was an easy test for our instrument. Here are some pngs from SAV: I presume the error rate and % perfect reads were derived from the ~1% phiX that I spiked in. -- Phillip

11-20-2012, 06:18 AM	#2
yaximik Senior Member Location: Oregon Join Date: Apr 2011 Posts: 205	Could you post also BioAnalyzer electrophoregram, the intensity curves (all Ns, or only C) over the entire run and thumbnails for C at cycles 30, 160, and 240?

11-20-2012, 06:59 AM	#3
NextGenSeq Senior Member Location: USA Join Date: Apr 2009 Posts: 482	This is similar to what we see.

11-20-2012, 08:37 AM	#5
yaximik Senior Member Location: Oregon Join Date: Apr 2011 Posts: 205	Thanks. This basically confirms what I suspected from my shorter runs. The usable length of fragment is between 1/3 and 1/2 of total length as descending polymerase molecules begin to crowd and interfere with each other in clusters. Or substrate diffusion becomes a limiting factor, or may be both. On shorter fragments this is more visible as cluster numbers begin to drop drastically and intensity curves fall off the cliff. Drop in intensity eventually generates error and aborts the run. In your example bell curves already began to form flattening over their maxima, which are very broad likely due to larger library size heterogeneity. So to take full advantage of 2x250 kit one needs isert size no less than 700 nt, better longer. Perhaps optimal cluster densities with such insert lengths need to be smaller.

11-20-2012, 08:55 AM	#6
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	There are lots of possible factors that contribute to worsening results as the number of cycles increases. Polymerases interfering with each other as they approach the flow cell surface seems a little far fetched. Also, generally one gets somewhat higher quality with shorter amplicons on Illumina instruments. So that doesn't really fit your hypothesis. -- Phillip

11-20-2012, 11:29 AM	#7
yaximik Senior Member Location: Oregon Join Date: Apr 2011 Posts: 205	Certainly. The hypothesis is what it is - a hypothesis, more data is needed to prove or disprove, but it seems explaining better what I was observing in my runs than ILM techsupport did. I want to prepare longer libraries with narrower fragment distributions and test the hypothesis more.

11-20-2012, 02:35 PM	#8
HESmith Senior Member Location: Bethesda MD Join Date: Oct 2009 Posts: 509	I'm in agreement with Phillip - after all, adapter dimers yield great quality scores...