Guardant and Twinstrand

So it seems Twinstrand are suing Guardant over use of their duplex sequencing approach. This suggests that Guardant’s digital sequencing and Twinstrand’s have similarities.

I’ve previously written about Twinstrand and posted a few further thoughts for paid subscribers. This latest litigation is interesting because it pulls a few things into focus.

Firstly, one of the issue I had with the Twinstrand approach is that you end up throwing away so much sequence data. In one example, only 3.5% of duplex reads had a pair present. This is a massive throughput hit. It’s difficult to see how this would be usable for whole genomes, or total cfDNA.

It makes a lot more sense with a targeted approach like Guardant360. In their 2015 paper. Guardant target a 78Kb region, and obtain a 8000x coverage using their digital sequencing approach.

Guardant also use “a duplex library whereby each single-stranded half of the original double-stranded input cfDNA sample is separately encoded with said oligonucleotides” and show a “typical error rate of approximately one error per 3,000,000” which is ~Q60. This all sounds very similar to Twinstrand.

Assuming Guardant have a similarly poor throughput of paired reads, they would need 62.4Gb to sequencing this targeted region. This is pretty tractable, and should cost a few hundred dollars using a Novaseq.

Guardant appear to charge ~$4000 to $5000 for Guardant360, so this seems acceptable.

The question is then, is do Guardant use an approach similar to Twinstrand?

While their papers are vague, the approach shown in Guardant’s patents looks fairly similar to me. As with Twinstand’s approach, Guardant show Y-adapters creating an asymmetry between the forward and reverse strands of dsDNA. This allows them to determine if a mutation was present on both strands or was introduced during a later amplification step:

In one of the Guardant examples they discuss “bubble-containing adapters”. These seem similar to Twinstrand’s Y-adapters, but they come back together:

While the approaches seem similar, it’s difficult to know exactly what Guardant have been using in their commercial offering. Litigation may reveal more details of Guardant’s approach, and will tell us if they’ve done enough avoid IP issues.

But this may already help explain why Twinstrand was formed and how it’s been able to raise $73.2M.

Twinstrand was founded on Oct 21st 2015, only 6 days after Guardant’s digital sequencing paper was published. It seems reasonable to assume that Guardant success with “digital sequencing” helped motivate Twinstrand’s foundation and financing.