NextSeq500 optimal cluster density?

[austinso]

Starting using our NextSeq500, and trying to gauge what the consensus is for ideal cluster density. FAE recommends 150-200K/mm2.

More importantly, I wanted to know what people's experiences are about the correlation between library concentration and cluster density and what people have been able to dial in.

Best

Austin

[fanli]

Short answer is it's complicated

We've run a bunch of different libraries (RNA-seq, ChIP, metagenomics) using different kits and see wildly varying results re: loading concentration and cluster density. Most of our RNA-seq libraries are either Clontech+Nextera XT or Truseq, and those seem to behave well with regards to cluster density. 150-200K/mm2 sounds about right - we have gone up to ~280 before with not a terrible drop in quality.

Where we've seen a lot of variability is with shotgun libraries using Nextera XT. I think our last run we loaded 5x the recommended concentration to get a density of 131. Then we heard from Illumina that Nextera XT wasn't really intended for use on NextSeq...but that's cool

If interested, I can ask the bench people and get actual loading concentrations they've optimized.

[GW_OK]

We shoot for loading at 1.6pM. It seems to work very well for us. We've gone up to 1.8, and down to 1.4, but we usually go back to 1.6.

Here are some quick graphs from 40 past runs showing how cluster density affects %PF, %Q30, and total reads. This data is Lane 1-Read 1 and is a mix of Mid and High runs with various read lengths and library types.

You can see the almost linear relation between cluster density and %PF (and obviously total number of reads). Cluster density doesn't seem to affect %Q30 in any meaningful way. It's generally library specific, I think.

[austinso]

We shoot for loading at 1.6pM. It seems to work very well for us. We've gone up to 1.8, and down to 1.4, but we usually go back to 1.6.

Here are some quick graphs from 40 past runs showing how cluster density affects %PF, %Q30, and total reads. This data is Lane 1-Read 1 and is a mix of Mid and High runs with various read lengths and library types.

You can see the almost linear relation between cluster density and %PF (and obviously total number of reads). Cluster density doesn't seem to affect %Q30 in any meaningful way. It's generally library specific, I think.

Thanks! Interesting that it seems that q30 seems to collapse at ~200K/mm2 with the exception of a couple of "outliers".

Question: There seems to be two kinds of sequencing kits or 2 library chemistries here, looking at the PF vs cluster density plots (last one). Are those 8-9 libraries at the lower PFs part of the group that tail-off in the Q30 plot (2nd one)?

Best

Austin

[Markiyan]

Please take into the account insert size distribution!

Dear GW_OK,

Thanks you very much for a charts,

It would be nice to make some sort of the 3D plot with average library insert size, and see what sort of surface do we get.

[GW_OK]

Those would be the Mid kits, which intrinsically have lower throughput.

Going back and looking at the data more closely, the "collapse" in Q30 seems to be mostly the Mid kits.

[austinso]

Huh...interesting! Very much appreciated!

Do you have an idea of what those 2 really high cluster density runs/samples were?

Best

Austin