phi x control on the same lane with sample?

[jasmineja]

We would like to run RNA seq- we have 8 samples. Do you have any experience with running the phix control WITH one of the samples? Could we still use the phi x as the control this way?

[Bruce E]

mRNA seq

We have been running RNA-seq with out a PhiX for over a year and it works fine. Giving up a lane for a standard is too expensive. The only reason you need a PhiX besides QC is if you are doing an unbalanced sample such as ChIP-seq, small RNA, Methyl-seq, or an unbalanced sequence. If you really want you can spike PhiX into your your sample at ~1% for QC but I do not believe this would help for phasing or matrix generation if your sample is unbalanced.

[jasmineja]

Thanks for your reply, the possibility of running with no phix is appealing.could you specify the description of balanced or unbalanced samples,
thanks

[huguesparri]

It's just that the GAII "calibrates" its matrix for the base calling step in the first cycles. In order to have a good calibration, the base abundancy in these cycles must be rather equal (25% A, 25% G, 25% C, 25% T, more or less).
Usually, you use PhiX as a reference balanced sample. But if you know that at least one of your samples is balanced, you can specify to the machine to use it as a reference.

[dhinerfeld]

If you do run into issues with a run, Illumina is much more likely to replace your reagents if you have run the phiX control. Just a warning.

[cbrennan]

As a core lab - we run a .04pM PhiX spike in each lane. Even if the clients library fails to produce clusters there is still enough to align PhiX & show that it was not an instrument problem. Bruce is correct it won't help with matrix/phasing etc if you have a biased sample you're trying to rescue.

The version of RTA/OLB you're using makes a huge difference, the newest RTA (1.6.47) handles RNA & CHiP much better than previous versions. If you do need to use OLB & a control lane to rescue samples - you don't have to use PhiX. Say you've got a good human genomic lane on that same flowcell -- you can use that as your control-lane. The only situation I've come across that you can't rescue is when the lane is just too over-concentrated - but a control lane wouldn't help in that situation either.

[mattanswers]

The version of RTA/OLB you're using makes a huge difference, the newest RTA (1.6.47) handles RNA & CHiP much better than previous versions.

I am working with ChiP and have GAPipeline 1.4.0. Can I get the newer version from Illumina ?

[cbrennan]

Yes -- you need an iCOM account, I believe you need your FAS to set it up for you.

Go here:

https://icom.illumina.com/

Once it's set up there will be a download section -- look under software.

You can download OLB v1.6.1 and run basecalling from images -- but it sounds like you're perhaps still on the 1.44mm flowcells, so if you're going to use Gerald for alignment you may have to use Casava v1.5.1

[mattanswers]

cbrennan,

Thank you very much for the information. Actually, we have been getting our sequencing done at various places so I am not sure of the flow cell size. I guess it is something I will have to pay more attention to.

[jasmineja]

thanks for the idea of running 0.04 pM phiX in each lane.
I installed the newest RTA, anyway it is neede for v.4.so we are ready to run 8 samples instead of 7... we will try. thanks

[cmawhinney]

As a core lab - we run a .04pM PhiX spike in each lane. Even if the clients library fails to produce clusters there is still enough to align PhiX & show that it was not an instrument problem. Bruce is correct it won't help with matrix/phasing etc if you have a biased sample you're trying to rescue.

When do you add your PhiX to the sample? I'm assuming if it's already at a pM concentration then it's after denaturation and you're adding PhiX diluted in HT1 to your samples that are denatured and diluted in HT1 as well.

Also we're multiplexing samples (running a control lane) but wanted to spike our samples with phix to help give the sample lanes more diversity. Does anyone have any experience with that? I'm wondering at what concentration to use. A little background, our barcodes were designed ourselves, we are not using Illumina's barcoding system.

[btarrier]

PhiX Spiking

When do you add your PhiX to the sample? I'm assuming if it's already at a pM concentration then it's after denaturation and you're adding PhiX diluted in HT1 to your samples that are denatured and diluted in HT1 as well.

Also we're multiplexing samples (running a control lane) but wanted to spike our samples with phix to help give the sample lanes more diversity. Does anyone have any experience with that? I'm wondering at what concentration to use. A little background, our barcodes were designed ourselves, we are not using Illumina's barcoding system.

We add PhiX to our samples after denaturation during the final dilution. We denature PhiX and dilute to 4pM as you would any other sample. Instead of making the final sample dilution in 1ml, we make it up in 990ul and then add 10ul of 4pM PhiX. This gives you a final concentration of 0.04 pM PhiX. This is typically about 0.5% depending on the sample concentration that we are loading.

I don't think that this will help you at all with base composition in barcoded samples. The software is looking for a somewhat uniform base distribution and 0.5% will not help if you only have two (or three) out of four bases present on any given cycle. We also use our own barcodes and PhiX at that concentration is basically invisible compared to the samples present.

[KorNor]

I use the same sort of protocol as btarrier.
Denature sample and PhiX separately, add PhiX to sample. However, I use only 2uL of a 600pM PhiX (subtracting that 2uL from the HT1 volume added to the sample), giving 1% PhiX.
So far this has been working well.
We simply include a balanced sample in each run and set it as the control lane, thus avoiding sacrificing a lane to PhiX
We've never run a flowcell without an unbiased sample, as the addition of PhiX is not enough to normalize the base ratio and assure correcting calling.

[cmawhinney]

I don't think that this will help you at all with base composition in barcoded samples. The software is looking for a somewhat uniform base distribution and 0.5% will not help if you only have two (or three) out of four bases present on any given cycle. We also use our own barcodes and PhiX at that concentration is basically invisible compared to the samples present.

We ended up spiking with 30% PhiX (at 7.5pM) because we realized spiking in such a low % of PhiX wouldn't help. Doesn't look like the 30% helped much either since Pipeline is having a hard time with base calling.

These samples were run with v2/3 reagents and thus run on version 1.5 of the Pipeline. I am now trying to run on OLB 1.6 as well as tweaking a lot of the parameters (for example setting it to use 10 bases, as opposed to the default of 4 in OLB 1.6) for cluster identification and it's not helping that much. FYI, We did run a separate control lane of PhiX as well.

If a anyone has any tricks for the analysis of barcoded samples! Since front indexing isn't an Illumina supported protocol they aren't much help.

[cbrennan]

cmawhinney -- did you try running OLB on a few tiles out past the barcode to generate a crosstalk matrix, and then run a full pipeline from cycle 1 importing that matrix? That's worked for me in the past.