A 2022 NGS Cost and Throughput Comparison

With Singular Genomics announcing the new G4 system at the end of 2021, and Element Biosciences announcing the AVITI system in March 2022, here is a look at the existing state-of-the-art systems in a single graphic

After spending some time in Los Angeles this week visiting a few existing and potential customers for Olink Proteomics, I was able to talk with a few friends about their perspective at this particular juncture in the NGS sequencing field. One of the big questions we discussed is what is needed to push the entire field forward, and a consensus is that it is not in the cost and throughput area between an existing Illumina NextSeq 2000 and an Illumina NovaSeq, but rather above the throughput of a NovaSeq. At an existing throughput of 6 terabases (at 90 gigabases per 30x whole-genome sequence, that’s about 67 WGS) in a little less than two days, well that’s the bar that has been set.

In the meantime I’ve been told that Singular Genomics is garnering interest, and if you are interested in details about their G4 I can refer you to their website and to James Hadfield’s post about the Singular G4 here, where he digs a little into the platform performance. Of course with four independent flowcells and four channels per flowcell there is some flexibility there. However for a compelling reason to adopt a new system the dimensions of differentiation often boil down to three important parameters: quality (both quantitatively and qualitatively), cost and throughput.

Quantitative and qualitative dimensions of NGS quality

Quantitative aspects of quality are clear enough: NGS accuracy is measured with Q scores, a simulated measure of Sanger Phred scores. Way back in 2014 and 2015 I wrote up a few posts for TMO’s Behind the Bench blog where I presented the history of Phred in detail, and wonder of wonders it still lives on. Suffice it to say it is defined as Q = -10 log₁₀(P), where P = probability of an error and Q=Phred Quality Score. Thus a Q30 is a 1 in a 1,000 error, or an error rate of 0.1%; at Q40, that is a 1 in a 10,000 error, or an error rate of 0.01%. Thus the measurement of all the bases on a base-by-base measurement, counting all the many Gigabases of sequence and determining the percentage bases above Q30, or the percentage of bases above Q40, will give that quantitative measurement of NGS quality for a given platform.

For qualitative dimensions, yes that is read format, and for the long-read, single-molecule market (dominated by Pacific Biosciences and Oxford Nanopore) there are long reads (many kb to 10’s or 100’s of kb) at the cost of accuracy. Refinements continue to be made, but at present there are real tradeoffs in accuracy and throughput when going the long-read route.

So given the majority of the market is in the short-read platforms, will keep this comparison between the two announced platforms from Singular and Element; the Pacific Biosciences’ Omniome short-read platform will not launch until 2023.

Presenting a simple chart comparing NGS cost and throughput

I credit a fair amount of this data to Twitter user Albert Vilella’s chart of NGS platforms as a Google Doc spreadsheet. Of course I had to correct a few errors, change the order and presentation and redo some of the calculations, so here you go.

This table shows how the Element’s cost-per-Gb spans the gap between a desktop system and the NovaSeq. And what is interesting are rumors of MGI launching a presence in North America and Europe in 2023 or perhaps even earlier. I haven’t heard anything directly about the T10 (one friend referred to it as a “T7 Plus”), and the numbers are from Albert’s spreadsheet (and is not sourced) so take it for what it’s worth, something interesting to think about, another 5x reduction in the per-gigabase cost.