At one level, Element's AVITI24 and Singular's G4X approaches are very similar. Both require an upgrade to the standard instrument, which as I understand it in both cases is purely an upgrade in compute power - optics and fluidics are untouched. So these should appeal to core labs wanting flexibility and bench-space savings and even larger labs that also like high reliability that comes with flexibility.
Both operate using special flowcells customized for spatial applications, with options of either a single huge field or multiple sub-compartments. Both offer fluorescent equivalents of H&E staining to visualize cell morphology. Both promise to simulataneouly assay a focused panel of DNA, RNA and proteins. Both are still early stage; Element plans to launch in the second half of this year and Singular saying "the end of this year".
But in details, we see differences.
First off, there are the two companies. Element has been more successful in the marketplace, with around 150 AVITI instruments placed - primarily in the US but some European and Asian placements. Singular was hobbled early on by supply chain issues and seems to still be under 20 instruments placed.
Second, there is company stability. Element is private, so there is no way to know how much money their treasure chest still holds, but Singular is public and as of their last quarterly report (in September; another must be coming very soon) they had about 10 quarters of cash left if you just naively divide cash by quarterly loss. That's a very, very simple model and with holes in each direction, but about the level of financial sophistication I tend to work at. Anything fancier requires free parameters to estimate, and as a colleague at Millennium said at an internal workshop once said "why take the effort to make up many numbers when you can only make up a few?". Singular did report an unaudited quarterly revenue number of $1M at the end of 2023, which is twice the previous value but very low when the instruments go for about a third of that. Element is probably far from breakeven, but even modest reagent sales per instruments (say $20K/quarter) times 150 instruments could easily beat that even before adding in instrument sales.
But perhaps most striking is the difference in markets being targeted. Singular is going after the obvious market of fresh/frozen tissue and FFPE, the workhorses for late stage translational research and possible future diagnostic use. They even have this little precision device to assist with mounting tissue onto the flowcell. That's where essentially all the incumbent spatial players are going and is a demonstrable market. Back at Infinity, we would have been champing at the bit to access spatial technology, particularly when our hedgehog inhibitor program had a tissue-based therapeutic hypothesis in pancreatic adenocarcinoma - our drug would modulate the tumor stroma (non-transformed tissue surrounding the tumor) to allow chemotherapy to better access the tumor. Even if we didn't have essentially a spatial hypothesis, there was the fact we would have a small bit of FFPE material to get maybe 10 slices, and from each we could ask a single question by IHC or FISH (or my crazy project, by Illumina sequencing). And of course no two slices through a real tumor are really equivalent! So Singular is going for an already defined, easy to understand and therefore already crowded market.
Singular's toolkit for mounting tissue onto the flowcell
In contrast, Element with their "Teton" chemistry is going to a much less already defined market for spatial omics: tissue culture. Element's system doesn't accept tissue, but instead an experimenter will grow cells in the "Teton" flowcells and then read them on the instrument. Element is betting that early stage drug researchers will consolidate multiple assays, such as the annoyingly-named "In Cell Westerns" (if it has neither a gel nor a blot, how is it a Western?) , into AVITI24 assays. They are also betting on use of AVITI24 for CRISPR-driven screens such as Perturb-Seq, with the spatial and cell imaging information adding new possibilities for defining phenotypes. This hasn't been the focus of extant spatial companies and more the purview of Berkeley Lights (which became PhenomeX which was acquired by Bruker) and startup Flexomics.
Element's Teton workflow for cell profiling
I would strongly argue that in the space "content is king" and that this will ultimately drive adoption or non-adoption, with content here meaning the mix of DNA probes to detect specific nucleic acids and affinity reagents (most likely antibodies) to detect proteins or protein states of interest. As noted in a recent Rahul Satija preprint comparing incumbent spatial platforms, analyzing multiple analytes in a spatial context can result in interference. Or go back to the original George Church parallel polony sequencing paper, which noted that maximum sequencer performance would be achieved if every imaging pixel delivered a different base. With direct spatial imaging, the number of pixels within a given cell is limited and may be less than desired to read all the analytes of interest with sufficient dynamic range, and conversely many pixels of the image will be wasted on the spaces between cells.
Both instruments are proposing to detect <100 analytes, so they must be chosen carefully to appeal to core problems of great interest in biopharma, and with appropriate abundance / affinity so that interesting minor components aren't crowded out by less interesting highly abundant ones. In other words, think carefully about using GAPDH as a control if you want to assay low abundance phosphosites.
So it's going to be an interesting and on-going challenge for these companies to build and validate panels. For some of these, the panel may just give yes-no answers, but for others actual focused sequencing will be possible - Singular talks of reading out immune repertoire diversity and KRAS codons 12 and 13 are candidates for a probe. But it is likely that many such panels must be developed, and actually assessing spatial multiomic data quality is not trivial, as the Satija preprint and two others referenced in it (Cook et al & Wang et al) demonstrate - and those were all looking at RNA and not the DNA-RNA-protein trifecta.
There's also the question of IP; 10X Genomics has been suing about every other player in spatial omics and has been frequently winning. 10X drove 1Cellbio into liquidation and on the eve of AGBT drove Nanostring into a Chapter 11 bankruptcy, which Nanostring has vowed to emerge from ("we're not going anywhere", as CSO Joe Beecham defiantly declared in a fiery Bronze Sponsor oratory). Nanostring quite likely won't be an independent company much longer, but if they find a deep-pocketed acquirer 10X may find them thinking about how sowing the wind reaps the whirlwind. But 10X has certainly been aggressive, and part of me is surprised that Element and Singular weren't served papers during their AGBT Sponsor presentations. Kidding there. But given that 10X bought 2020 AGBT sponsor ReadCoor which held extensive IP on true spatial sequencing (FISSEQ invented by George Church's lab), the probability that Singular and Element at least approach 10X's litigation Schwarzchild radius seems high.
How will it all play out? That's the fun part of this business -- nobody knows! Spatial genomics has been a bit of a wild field already, with many startups entering the market, a number of companies already fading out and a multitude of different sub-technologies at play. For example, if you are detecting proteins is that via conventional secondary antibodies, direct fluorescent labeling, nucleic acid tags or mass spectrometry tags? Is spatial information captured by actually imaging a labeled molecule in location, or by encoding the spatial position in nucleic acid tags to be read on a conventional sequencer. Is the readout of nucleic acids reading tags or building up a tag by sequential probing, or is the tag (or a native molecule) actually being sequenced?
I'm still trying to build a complete scorecard of all the players so I have some hope of tracking the field. But a census of the players is just a start - what metrics does the market really care about here - capital cost, operating cost, resolution, plex, throughput, size of image field, etc and how do the different platforms perform for these on different sample types? Makes the traditional NGS field look simple!
[2024-02-29 00:05 - fixed "Trinity" to "Teton" in Element section - Teton in the Element spatial chemistry; Trinity is hybrid capture]
No comments:
Post a Comment