no index reads

[Tobias Kockmann]

Hello HiSeq2000 users,

we just finished a multiplexed mRNA-seq run on the HiSeq2000 with puzzling results, maybe somebody encountered this phenomenon before:

We prepared 6 mRNA-seq libraries using the illumina TruSeq kit incl. the indexed RNA adapters and sequenced using a 50x cycle single-end run on the HiSeq 2000 on a single lane. The insert cycles (1-50) yielded sequences of reasonable quality, although the cluster density on that lane was pretty high (around 1'000 clusters/mm^2). From cycles 51-58 (the index/barcode) we do not get any sequence information (NNNNNNNN), which means the do-convolution of the samples failed. On the same flow cell there was another multiplexed sample, using the same indexing system (illumina TruSeq) and this worked out pretty fine.

Any idea, what went wrong here???

[pmiguel]

Never had this issue with 6 indexes in a lane -- usually only a problem when there are less than 4 indexes for us.

My speculation is that one or two of the six samples produced far more clusters than any of the the others. Under such circumstances I find the instrument frequently fails to obtain sequence of the indexes.

I blame this on low diversity of sequence during scanning. But thus far I don't seem to be convincing many people -- especially at Illumina tech support.

Are you sure that all 6 bases were unreadable? I just found out about a parameter:

--use-bases-mask

that can be used to "turn-off" failed bases in the bar code and force demultiplexing base only on the bases that work. Well, I may be making sound more sophisticated than it is. You have to determine which are the bad bases yourself (or with the help of tech support) and then feed the switch above a string that directs it not to use certain base numbers.

--
Phillip

[Tobias Kockmann]

Unfortunately all of the index cycles failed to deliver sequences, which means I can not use partial information to demultiplex the samples.

Never the less, thx for the hint!!!

[GenoMax]

So only your sample failed to generate sequence for the index read cycles whereas another set of samples on the same flowcell worked right?

Did all your reads end up in "unknown" directory after de-multiplexing?

Unfortunately all of the index cycles failed to deliver sequences, which means I can not use partial information to demultiplex the samples.

Never the less, thx for the hint!!!

[Tobias Kockmann]

So only your sample failed to generate sequence for the index read cycles whereas another set of samples on the same flowcell worked right?

Did all your reads end up in "unknown" directory after de-multiplexing?

Exactly, for my libraries we didn't get any sequence information from the index, but there was another multiplexed sample on the same flow cell that worked without any problem. Since, all the lanes on the HiSeq 2000 share the same reagents for sequencing, I consider this as a sort of positive control, which pinpoints to a problem which seems to be library specific.

[Tobias Kockmann]

...and yes, all of my reads ended up in the "unknown" directory after demultiplexing! Without exceptions!

[GenoMax]

This sounds like a problem that is specific to your samples then since the other samples are working fine. Guess you could try running your samples again to see if this was a one time issue.

Exactly, for my libraries we didn't get any sequence information from the index, but there was another multiplexed sample on the same flow cell that worked without any problem. Since, all the lanes on the HiSeq 2000 share the same reagents for sequencing, I consider this as a sort of positive control, which pinpoints to a problem which seems to be library specific.

[Tom Barker]

Hi,

I have recently had a similar problem. The samples all clustered at ~600k/mm2 and the index reads were very good quality according to SAV however when the data was passed onto our bioinformatics team two of the lanes gave no sequence information for the index reads. I was just wondering if you ever got to the bottom of this.

[pmiguel]

One thing to check is whether you get any demultiplexing using all 24 TruSeq v2 indexes. If, for instance, the wrong adapter got added to a library, everything would go into the unknown directory.

--
Phillip

[Jon_Keats]

Can you look at the on machine base composition stats for the index read? If so do you see the typical craziness or a flat line at zero for this lane in question? As suggested I'd check that you didn't submit the wrong indexes to the demultiplex step so maybe try submitting all 24 or double/triple check the library prep to ensure it is right.

[HESmith]

Using the wrong indices for demultiplexing would not produce Ns for the index read (see the first post). We've seen a similar phenomenon (good read one but no index bases called), but only at high (1000K) cluster densities. Any chance that the sample was loaded at too high a concentration (pmiguel is right that the reported densities are inaccurate in this situation)? The %PF would be lower than normal, and the thumbnails would show the actual cluster density.

[pmiguel]

Using the wrong indices for demultiplexing would not produce Ns for the index read (see the first post). We've seen a similar phenomenon (good read one but no index bases called), but only at high (1000K) cluster densities.

Actually, this is a concatenation of 2 threads. The first was started by Tobias Kockmann in September of last year. The second was started by Tom Barker a few days ago. Tom's results are qualitatively different than Tobias's. Tom has high quality values and a density of 600K/mm2. I am presuming that is his raw cluster density, though he does not specify.

--
Phillip

[Tom Barker]

Hi,

I was originally told that the index reads only produced N's but have recently found out that they did produce sequence and as it turned out it was the wrong indexes. Could have saved me alot of time looking into this if he had of said the indexes don't match as oppose to they're all N's. Thanks for all your input anyway.

Tom

[pmiguel]

I have never seen sequence reads generated during demultiplexing. Perhaps there is a switch to make that happen? Would be handy in cases like these.

Anyway, that is why if I see there was "no information" from the index reads, I interpret it as "none of the indexes I specified were found during demultiplexing".

--
Phillip

[cbird]

possible explanation

I'm a novice, but I think that high cluster densities could lead to successful reads of your target DNA and unsuccessful reads of your indexes.

The Illumina sequencers use the first few sequencing cycles to identify clusters, and I believe this happens on the fwd read, the rev read (paired end), and the index read. There are only 6 unique signals in an index read (if six indices are used) versus 4^4 potentially unique signals in the target DNA reads. Consequently, high cluster density could lead to deconvolution problems. That could explain the pattern the original poster reported - that is, of course, if I understand what the sequencer is doing.