When I originally reviewed Singular Genomics I assumed they were using random clusters. This would have been pretty much the same approach as Illumina used in the original Genome Analyzer, and currently on the MiSeq.
Of course, this approach severely limits throughput. But it has the advantage of being based on IP that has long expired. However someone pointed me to the Singular website, which shows something that looks very much like a patterned surface:
After a little more hunting I found an explicit statement that “we use patterned flowcells in the G4”. This suggests they are doing bridge amplification on a patterned surface, as Illumina did in the HiSeq X and later instruments:
Other public statements also state that “High-density clustering and clonal amplification of Max Read libraries on a patterned flowcell allows multiple template molecules to amplify within the same nanowell of the array”.
Elsewhere in their S-1: they state that “two steps of the sequencing process occur on the flow cell, which include cluster amplification and the sequencing chemistry”.
James over at enseqlopedia however dug up a patent that describes a bead based cluster generation approach. This patent discusses cluster amplification on beads, which are then loaded into an unstructured flowcell creating a 3D array of beads which can be imaged by confocal microscopy. It sounds pretty cool. It also sounds very complicated and difficult to get working at high throughput!
So I went hunting through a few more patents. And came across this one. I suspect this may be the approach being used (though I’d love explicit confirmation!).
In this patent, cluster generation happens on beads loaded onto a patterned flowcell:
Clusters are then grown on these beads and imaged fluorescently. Images presented in the patent look similar to those shown in public presentations:
The question is what advantage does this confer? James suggested that the approach may avoid ExAmp and associated index hopping issues.
To me it seems like a way of avoiding doing bridge amplification directly on patterned flowcells (which is what Illumina do). Looking through Illumina’s IP I imagine this is covered in this patent which doesn’t expire until 2032.
If so this is a pretty neat way round the problem!
If you have any further insight please get in touch (new@sgenomics.org) or share your thoughts on the Discord!
Interesting. This looks like a fair bit of work, so thanks for taking the trouble!