Roche will have a presentation on the 20th of February to give more details about their Xpandomer-based SBX sequencer.
From the abstract:
“Sequencing By Expansion (SBX) technology is a novel sequencing approach that encodes the sequence of a target nucleic acid molecule into a highly measurable surrogate polymer called an Xpandomer. Xpandomers encode the sequence information into high signal-to-noise reporters, enabling high-fidelity, single-molecule nanopore sequencing using a highly parallel 8M sensor array to generate throughputs of over 500M bases/second”
Roche is bringing this new SBX sequencer after many years of polishing the technologies they brought in-house from the acquisitions of Genia and Stratos Genomics. The Genia acquisition, which took place in 2014, now 11 years ago, made people nervous about how serious Roche was about NGS. Later on, in 2020, Roche acquired Stratos Genomics. This brought in the SBX (Sequencing by Expansion) technology which was then combined with the Genia technologies into a single product that is going to be announced in a month’s time.
At the time, Stratos was touting a technology that would produce reads of an average of 500bp, with more than 90% above Q30.
Roche since then presented about the Xpandomer chemistry, including the key benefits of great signal-to-noise (SNR) ratio to produce highly accurate reads.
This would be for an 8 million microwell chip, each sequencing individual expandomers in parallel.
What the abstract alludes to is a system producing over “500M bases/second”, which means that it would produce 1Gbp in 2 seconds and it would take 2x8000=16,000 seconds or 4.4 hours to produce the 8Tbp of data that the Illumina NovaSeq X 25B or the Ultima Genomics UG100 produce in 24 hours.
If the reads were of around 500bp, this would match well with the kind of read pair cycles that Illumina does (2x150=300bp for the most affordable price per Gigabasepair, and 2x300=600bp but pricier), and Ultima Genomics in their 250-350bp single-end read range.
Now let’s talk accuracy. The basis for accuracy in NGS is “how many nucleotide errors per N length does a technology have?”. The standard Q30 is 1 error per 1,000 bp, which means that Illumina 100bp reads would produce 9 perfect reads and 1 read with 1 error in the 100bp sequence, on average, with a Q30 score. In the long-read technology front, Q30 would mean that an Oxford Nanopore 1,000bp read would on average have 1 error, and 10,000bp reads would on average have 10 errors sprinkled along the read length. Oxford Nanopore has nowadays a modal Q25 for simplex reads, which means that there are on average 3 errors per 1,000bp molecule instead of 1. On the other hand, Oxford Nanopore duplex reads are nowadays in the Q30 range, so apples-to-apples similar to Illumina and other short-read technologies out there.
https://en.wikipedia.org/wiki/Phred_quality_score
The dark corners of the NGS world, the kind of “everything-goes” caustic forums where people only dare to post anonymously, have speculated about what the Q-score of the Roche SBX sequencer could be (I read these forums so you don’t have to). There seems to be a large number of people aiming at Q20 as their prediction, which equates to 1 error per 100 bases. So if the average read length for the Roche SBX sequencer is 500bp, this means it will produce reads with 5 error sprinkled in each 500bp sequence.
Strategic Positioning
Let’s suppose the new Roche SBX sequencer is a $1M machine that produces 500bp reads of Q20 quality at a rate 4-6x faster than the Illumina NovaSeqX 25B and the Ultima Genomics UG100. What does this mean for the field of NGS? I’ll give my opinion below.