illumina library preparation and multiplexing

[jgibbons1]

Hi all,
In the next month or so I will be illumina sequencing genomic dna and cDNA of 14 samples. The genome/transcriptome size I am working with is moderate (37Mb/18Mb) and I am going to try multiplexing gDNA and cDNA in the same lane using different adaptors. I am wondering 2 major things:

(1) Could anyone provide me with library preparation protocols for gDNA and cDNA? Are there any potential problems I should be aware of? What is the best way to QC samples?

(2) Has anyone tried multiplexing? What advice do you have?

Thank you in advance...your comments are greatly appreciated.

-John G

[greigite]

Hi all,
In the next month or so I will be illumina sequencing genomic dna and cDNA of 14 samples. The genome/transcriptome size I am working with is moderate (37Mb/18Mb) and I am going to try multiplexing gDNA and cDNA in the same lane using different adaptors. I am wondering 2 major things:

(1) Could anyone provide me with library preparation protocols for gDNA and cDNA? Are there any potential problems I should be aware of? What is the best way to QC samples?

(2) Has anyone tried multiplexing? What advice do you have?

Thank you in advance...your comments are greatly appreciated.

-John G

There's a great older thread on this that I can't find just now, but if you are using your own multiplex adapters, be sure that you have balanced representation of all 4 bases in the first 4 bases of your tag set. This is necessary because cluster identification is based on the 1st 4 bases, and if you have an over-representation of some bases, the clusters will not be reliably separated. I have to admit that I didn't know this when I designed my own 6 bp adapters, and the result was that my lanes only contained around 70% as many reads as other lanes on the same flow cell.

[jgibbons1]

This is incredibly insightful. I will be barcoding 2 samples per lane and I had actually planned to use something AT (AAAAAT) biased and CG biased (CCCCCT). I'll be sure to change these. I'm not sure I really understand how this would alter cluster identification...if you wouldn't mind could you explain this to me?

Thank you thank you thank you!

[greigite]

This is incredibly insightful. I will be barcoding 2 samples per lane and I had actually planned to use something AT (AAAAAT) biased and CG biased (CCCCCT). I'll be sure to change these. I'm not sure I really understand how this would alter cluster identification...if you wouldn't mind could you explain this to me?

Thank you thank you thank you!

Sorry, wish I could explain it, but it was a core facility manager who told me about the cluster identification issue and he didn't explain how it works. Hopefully someone else can chime in here.

[Jessica_L]

From what I understand, the image analysis software expects that bases should be represented equally in a stretch of sequence and it goes off that expectation to calculate the amount of crosstalk between fluorescent wavelengths (part of the reason to use PhiX as a control lane if you are sequencing something that isn't approximately 25% ACGT, like genomic DNA).

If you have one base (like A) over-represented in your barcode, the software will suppress some of that signal in order to boost what it interprets as an artificially low intensity for bases C,G,and T. In other words, the matrix estimates will overestimate the amount of crosstalk on base A (i.e. the software assumes that fluorescence from the other nucleotides are partially responsible for the unusually high intensity of A), and greatly underestimate the crosstalk on the other bases (i.e. fluorescence from A doesn't bleed into the intensity values for the other bases).

The result of that is that the software will conclude that the values for C,G, and T are all legitimate, but the value for A is artificially elevated through dye crosstalk. The result is the leveling effect I mentioned above (where the bases that are overrepresented in your barcode are artificially suppressed).

That's going to cause the barcode sequence to not be called correctly, so instead of AAAAAT you're going to see a lot of ANNNNT or something like that. If you lose enough of the barcode, you may lose the ability to discriminate between specimens depending on the barcode sequences.

[elaney_k]

Hi guys,

The cluster seperation/identification is carried out in the first 4 cycles of sequencing for SCS 2.6, I think previously it was the first 2 cycles, so if all your samples contain the same initial sequences then the algorithm will stuggle to seperate neighbouring clusters as it effectively "sees" them as the same because the bases identified are identical.. hence the reason for balancing out the base composition of the indexes.

[james hadfield]

Word on the greapevine is that Illumina are bringing out new adapters which will be full length and barcoded. You might want to wait for their slot at AGBT to finish.

[greigite]

Word on the greapevine is that Illumina are bringing out new adapters which will be full length and barcoded. You might want to wait for their slot at AGBT to finish.

Really? Do you know when these will be commercially available?

[snetmcom]

Word on the greapevine is that Illumina are bringing out new adapters which will be full length and barcoded. You might want to wait for their slot at AGBT to finish.

i didnt hear this. anything come from it?

[Simon Anders]

I'm surprised about #5, I though part of the reason the new SolexaPipeline software gives so much more clusters now is that they segment using all images now.

Another reason I've been told for the need of balancing the adapters is that the microscope camera gets confused if a base appears in none of the clusters in a certain cycle because then, the image is completely black and the machine tries to compensate by wrongly adjusting exposure time for the next cycle. But this is just a rumor, no idea whether it makes sense.

Simon