Comparing Illumina's sequencers

Updated: The cost per M reads and cost per Gb figures in the original posting were wrong - damned Excel operator error! I've fixed them, again. Thanks to Shawn for his comments.

I've been asked about the difference between the Illumina sequencer line-up so many times that I put together a spreadsheet to help the discussions. This is cobbled together from the Illumina website and there are no prices quoted, however I have estimated the £ per M reads and the £ per GB.

Illumina sequencer comparison: Before using the spreadsheet I'd suggest you ask yourself a question "what are you sequencing: genomes, transcriptomes, exomes or ChIP-seq?" The two major applications are genome sequencing where the cheapest per GB cost is probably most important, or counting apps like RNA-seq and ChIP-seq where the cheapest per M reads cost is probably most important. Exomes fall somewhere in-between and I am still not clear on whether I prefer long reads or more reads (probably the latter).

Want long reads: On MiSeq you can get up to 25M 600bp and on HiSeq 2500 rapid up to 150M 500bp reads, I can't wait to see assembled genomes from those data, I wonder how well they'll compare to a Yeast genome on Oxford Nanopore? 

Faster sequencing: Illumina took a huge leap forward on the HiSeq 2500 when we moved from an 11 day (more like 12-13) run to just 6 days. Clustering on V4 also dropped from 5 hours to 2. Of course data volumes went up, but the big news for labs like mine was the fewer and faster lanes; which added up to a much easier life in the lab. Look at the continued drop in run-time on the HiSeq platform - I defy and PI to argue that they seriously need data faster than "next week", speed is still an issue for some clinical applications, but I'm not sure we need faster sequencers in the lab.

Sequencer read-length and run times

How much for those reads: The pricing of Illumina kits is not publicly accessible so I've estimated costs on a per GB or per M reads basis. The NextSeq really is a good alternative if you are not going to do very large projects, or cannot access a HiSeq. The costs are higher, but the up-front capital is much much lower than a HiSeq 4000. Having the 2500 in my lab (and the 4000 soon I hope) does allow us to run some pretty massive projects, and a significant benefit is that the whole project can often be completed in a single run. We do a lot of sequencing for a lot of people in Cambridge so the high-output format on 2500 and soon the 4000 really makes sense for us.

Updated Sequencer costs: per M reads of per Gb

How many samples per run: Most users want their project to be competed as quickly as possible, and we a re now a long way from running a single ChIP-seq sample per lane! Many of our RNA-seq projects are coming in at 96 samples and being able to process everything in one batch through library prep and sequencing means turnaround times can be great - potentially 2 weeks from RNA-data, and we could conceivably run 96 exomes in under 7 days (Nextera capture on HiSeq 4000 PE75)!

How many samples per run?

PS: Enjoy the info but please accept that although I took the figures from Illumina's website they, or my copying, may be wrong and/or subject to change. Any mistakes here will be rectified if I can get round to it. Don't buy a HiSeq 400 or a NextSeq because of what you read here, talk to someone!