Optimal cluster density for ChIP-seq in HiScan SQ.

[fogi]

Hi everyone,

I am about to have SR run (for ChIP seq libraries) on HiScanSQ. I wonder what is the optimal cluster density for unbalanced samples like ChIP-seq libraries (a lot of the same DNA sequences) with a PhIX spike-in around 40 - 50%? Does anyone spike-in the PhIX control for ChIP-seq, instead of scarifying one lane for it?

[C.R.]

we do not spike our ChIP-Seq Libraries, but some transcription factor ChIPs may benefit from this. However, we are using the 40-50% spike-in for RRBS libraries which are really challenging when sequencing is performed on a HiScanSQ. We usually used a PhIX control lane, but in the last runs we omitted it. Although I am not sure if this was a good idea. It is hard to say if the quality of the runs would have been better in the end if we would have included a control lane.

[tomaszzbk]

spike-in of PhIX

Hellow,
My question relates to the sequencing of RRBS library on HiscanSQ machine.
When you perform spike-in for any RRBS library do you make any indexing of PhiX to better filter it out at the stage of alignment? When not do you find any way how to remove the reads of PhiX spiked in the library of interest.

Thanks for your help.

we do not spike our ChIP-Seq Libraries, but some transcription factor ChIPs may benefit from this. However, we are using the 40-50% spike-in for RRBS libraries which are really challenging when sequencing is performed on a HiScanSQ. We usually used a PhIX control lane, but in the last runs we omitted it. Although I am not sure if this was a good idea. It is hard to say if the quality of the runs would have been better in the end if we would have included a control lane.

[C.R.]

We bought the PhiX from Illumina. I am not sure, if there is an indexed one available. At least in the past this was not the case. This means that in fact we cannot exclude that some reads may come from PhiX contamination. Actually, I always filtered the reads for those which start with either CGG or TGG. Only a low percentage of reads in the PhiX Lanes fullfil this criteria. Next, in the past we used full Lanes PhiX as control Lane. Whenever I mapped these pure PhiX reads using the same conditions with bismark to human or mouse genome only a 1000 reads out of 40-60 millions of reads mapped at all. So it is a really small contamination in the end. On the other hand maybe the spike-in is not the best idea: some month later Illumina published their own version of the RRBS protocol. Here they do not recommend PhiX but say: 'It is highly recommended that you limit the cluster number for RRBS libraries to a 75–80% optimized cluster number for other genomic libraries, because of the unique CGG starting sequence in the first three bases'. This way also seems to work.