To review, Scale's single cell sequencing kits implement a combinatorial barcoding scheme in which fixation converts the cells themselves - or nuclei - into nanocontainers by covalently linking nucleic acids to the cell structure. It is then possible to use rounds of indexing, pooling and splitting to build a statistically unique index onto each cell.
In practice, the initial fixation step involves several rounds of washing, which doesn't scale well with many samples and can lose material - an issue for very tiny samples. My training is from The Church of Multiplexing (though it's polite to call him George) and a near commandment there is to pool as early as you can. The key ScalePlex reagent enables tagging the cells during the first fixation step, so samples can then be pooled prior to any washing steps - very much fitting the "pool ASAP" principle.
Scale takes pains to not call ScalePlex cell hashing. It's an analogous procedure, but cell hashing has relied on either antibodies to cell surface proteins or on tagged lipids that interpose themselves within the cell membrane. Both of these can dissociate during later pooling steps, muddling the cell indexing - much as index hopping on the flowcells deranges careful assignment of sequence indexes to samples. Antibodies also have the disadvantage of being specific to some branch of life; there's no universal protein antigen across all eukaryotes. ScalePlex does bear a oligo-T capture sequence, so it is specific for eukaryotic Pol II transcripts - those of us interested in bacteria never catch a break!
There are 96 separate ScalePlex barcodes. But those ScalePlex-labeled pools can then go into the next labeling plate that also has 96 barcodes, which are added by ligation. After pooling-and-splitting - and Scale provides a bit of plasticware that enables pooling by just inverting the labeling plate - a mid-level of 384 indexes is added. After another pool-and-split operation, 96 final barcodes are added by tagmentation. All of these steps are the standard Scale 3' mRNA kit; the only change for ScalePlex is to use that reagent during the first fixation step and then pool samples.
In theory you could put 96 pools of 96 samples into that first plate. Probably most people will not go so far and spread each ScalePlex pool across multiple level 1 indexes. The kit as distributed has a specification of indexing 125K cells, though with an expansion kit that can go to 500K cells. So distribute those 500K cells amount however many samples you have; trying to go for nearly 10K samples (where would you get them?) would mean only 50 cells per sample. Is that enough? Certainly not for any sort of complex tissue, but maybe there are regimes where having large numbers of samples and small numbers of cells per sample is acceptable. Scale discussed at AGBT this year a roadmap to increase the limit to 2 million cells. But the key point is ScalePlex puts that flexibility easily in the experimenter's hands - it will be practical to design experiments with replicates, timepoints and so forth which generate many samples, but which would previously have been too onerous to execute on the lab side.
For now, sample inputs are limited to fresh or frozen samples - FFPE is not an option. Cells or nuclei work as input - illustrating again the value of a broadly-reactive tagging reagent. The limit to 125K is basically due to a bottlenecking in the last step, but it is also possible to store the not-quite-finished library material not put through that step and finish processing it later. So, if a reviewer demands more technical replicates it's easy to generate them. In "barnyard" experiments, ScalePlex does not significantly change the <5% multiplet rate seen without ScalePlex.
ScalePlex will be $20/tag x 96 wells = $1920. The base kit is $8800. For data generation, Scale recommends 20K reads per cell, so perhaps an S1 or S2 flowcell. So a full single cell experiment is around my first year of graduate school stipend - that's just the nature of this field. But if you have an idea for screening compounds within a hit-to-lead program in pharma that would be enabled by this technology, those sort of costs are not out-of-the-question.
One of the many ideas for this space that I've let sit fallow too long is that technologies such as ScalePlex are creating the opportunity for new "bottomless pits" of sequencing - good experiment concepts that require sequencing capacity that could not be dreamed of in the recent past but which is now in our grasp. When I was actually collaborating on designing experiments in drug discovery, we always needed higher sample counts - replicates, controls, more cell lines or timepoints or conditions. Without the ability to run that, why run at all? And I have both very painful and less painful examples of where controls were missing (very painful) that could have caught errors early or where the correct controls did catch a major experimental problem (still painful, but much less so). It's very hard to have too many controls!! If a tool such as ScalePlex can enable processing much better sample designs, even with the 125K (or 500K expanded) limit on cells per specific experiment it will, I believe, drive further penetration of single cell RNA technology into pharma drug discovery laboratories - which will mean more FASTQs being generated more frequently - which is music to any computational biologist's ears!
No comments:
Post a Comment