Massively parallel digital transcriptional profiling of single cells

Massively parallel digital transcriptional profiling of single cells Grace Zheng Jessica M. Terry Phillip Belgrader Paul Ryvkin Zachary Bent Ryan J. Wilson Solongo B. Ziraldo Tobias D. Wheeler Geoff P. McDermott Junjie Zhu Mark Gregory Joe Shuga Luz Montesclaros Jason G. Underwood Donald A Masquelier Stefanie Y. Nishimura Michael Schnall-Levin Paul W. Wyatt Christopher M. Hindson Rajiv Bharadwaj Alexander Wong Kevin D. Ness Lan Beppu H. Joachim Deeg Christopher McFarland Keith R. Loeb William J. Valente Nolan G. Ericson Emily A. Stevens Jerald P. Radich Tarjei S. Mikkelsen Benjamin J. Hindson Jason H. Bielas

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Massively parallel digital transcriptional profiling of single cells 2017 Nature Communications 10.1038/ncomms14049 zheng2017-massively-parallel-digital-transcr zheng2017-massively-parallel-digital-transcr.txt, the project's staged copy 187809f5bfc1c43c970e2f235eab9959e06fcab5 cc-by genomics

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them 3' mRNA counting of tens of thousands of individual cells in parallel - and, separately, telling whose cells they are from the same reads Quotation withheld because this record is not cleared for redistribution. whether tens of thousands of single cells can be captured and barcoded without custom microfluidics or plate sorting, and at what multiplet, capture and conversion rates Quotation withheld because this record is not cleared for redistribution. a gel-bead-in-emulsion platform capturing ~50% of loaded cells across 8 parallel channels, with its multiplet rate, cross-talk and conversion efficiency each measured against a control of known composition; plus a de novo SNV method that resolves transplant chimerism without either genotype Quotation withheld because this record is not cleared for redistribution. method Quotation withheld because this record is not cleared for redistribution. transcriptomic heterogeneity that population-averaged measurement hides, and the droplet-level artefacts - multiplets, barcode cross-talk, technical variance - that appear once you measure cells one at a time Quotation withheld because this record is not cleared for redistribution. omics-profiling omics-profiling Quotation withheld because this record is not cleared for redistribution. cells are loaded far below one per droplet so that two rarely share a barcode; the same dilution that suppresses multiplets is what throws away the cells that are never captured, so the multiplet rate and the capture rate are two readings of one loading decision limiting dilution as the price of purity Quotation withheld because this record is not cleared for redistribution. synthetic RNAs at a known concentration are loaded in place of cells, so every difference between the expected and observed counts must come from the instrument - the variance that remains is technical by construction rather than by argument a droplet with no biology in it Quotation withheld because this record is not cleared for redistribution. a barcode carrying substantial human and mouse counts can only have held two cells, so mixing species converts an invisible failure into a countable one - and the count must then be doubled, because same-species doublets are equally likely and leave no trace two species as a doublet detector Quotation withheld because this record is not cleared for redistribution. the ~250 nt of cDNA that gives an expression count also carries sequence variants, so each cell has an expression identity and a genotype identity from the same reads; agreement between them is evidence neither could give alone identity read twice from one molecule Quotation withheld because this record is not cleared for redistribution. an observed barcode one Hamming step from known sequences is reassigned by posterior probability from base qualities and the overall barcode count distribution, above 0.975, with lexical order as the stated tie-break rather than a silent one posterior correction of a barcode without a whitelist match Quotation withheld because this record is not cleared for redistribution. the level of detection of the genotype method is given as 3% in Results and as 4% of 6,000 cells in Methods, for the same in silico mixing experiment, and the two are never reconciled Quotation withheld because this record is not cleared for redistribution. the same discrepancy, from the Methods side - K=1 is selected when the mixture fraction is below 4% of 6,000 cells Quotation withheld because this record is not cleared for redistribution. multiplets of identical or highly similar cell types cannot be detected by any of the measures used here, and the authors say more sophisticated methods will be required Quotation withheld because this record is not cleared for redistribution. a 15-20% error in the cell counts feeds directly into the capture efficiency, and the authors name it as a source of the variability in that figure Quotation withheld because this record is not cleared for redistribution. imaged multiple-encapsulation was 16% higher than the Poisson expectation, ascribed to subsampling error or to cell-cell clumping seen during the manual count Quotation withheld because this record is not cleared for redistribution. classification blurred the boundaries between T-cell subpopulations, with part of the inferred CD4+ naive population classified as CD8+ T cells Quotation withheld because this record is not cleared for redistribution. the transplant work is a proof of principle on BMMCs from two patients Quotation withheld because this record is not cleared for redistribution. the unexpected absence of monocytes after transplant is reported without explanation and left for further study Quotation withheld because this record is not cleared for redistribution. the GemCode gel-bead-in-emulsion barcoding platform, carried from genome haplotyping, structural variant analysis and de novo genome assembly into single-cell RNA counting Quotation withheld because this record is not cleared for redistribution. the two-genome mixture model with a Gibbs sampler and Stephens relabelling, carried from mixture-model statistics into deciding which transplanted cells are whose Quotation withheld because this record is not cleared for redistribution. a mixture of ~1,200 human 293T and ~1,200 mouse 3T3 cells loaded together, so that a droplet holding two cells leaves a two-genome barcode behind it eliminates: multiplets and barcode cross-talk being invisible - in a single-species run a doublet and a large cell produce the same data, and neither rate can be read · outcome: held Quotation withheld because this record is not cleared for redistribution. a cell titration across six different cell loads, recovering 1,200 to 9,500 cells eliminates: the 1.6% figure being a fixed property of the instrument - if the rate did not track the load, the mechanism would not be Poisson co-encapsulation at all · outcome: held Quotation withheld because this record is not cleared for redistribution. high-speed imaging of every GEM as it formed, with ~28k frames counted by hand, as a measurement of cell occupancy that does not pass through sequencing eliminates: the Poisson loading claim resting on the same data it explains - the imaging can disagree with the sequencing, and it does, by 16% · outcome: partial Quotation withheld because this record is not cleared for redistribution. ERCC synthetic spike-in RNAs at known concentration loaded into GEMs in place of cells, in two mixes and two dilutions eliminates: biological variation as a contributor to the measured variance, and the conversion efficiency being inferred from data with no known answer · outcome: held Quotation withheld because this record is not cleared for redistribution. droplet digital PCR on eight genes in Jurkat cells, against the RNA copy number measured in bulk from the same cell type eliminates: naked spike-in RNA standing in for a real transcript inside a real cell - ERCC molecules are not lysed out of anything, and could convert at a different rate · outcome: held Quotation withheld because this record is not cleared for redistribution. 293T-only and Jurkat-only samples run through the identical PCA, alongside the mixed ones eliminates: the clustering method itself as the source of the two clusters - a procedure that splits a pure sample would have produced the same picture from nothing · outcome: held Quotation withheld because this record is not cleared for redistribution. cell identity assigned a second time from SNVs alone, independently of any marker gene, and compared cell by cell with the marker-based assignment eliminates: the choice of marker genes as the source of the cluster identities - genotype and expression are different measurements and could have disagreed · outcome: held Quotation withheld because this record is not cleared for redistribution. in silico mixing of two donors' reads at 12 ratios from 0 to 50%, where the true label of every cell is known, with sensitivity and PPV computed against it eliminates: a detection limit asserted rather than measured - the floor is found by mixing downward until the method stops recovering the minor population · outcome: held Quotation withheld because this record is not cleared for redistribution. the same two donors mixed in vitro at 50:50, 90:10 and 99:1 before sequencing, rather than mixed computationally afterwards eliminates: in silico mixing as the reason the floor holds - reads combined by arithmetic never pass through a shared droplet, a shared PCR or a shared flow cell · outcome: held Quotation withheld because this record is not cleared for redistribution. 11 pairwise genotype comparisons between samples from the same individual and 15 between different individuals, computed before any patient sample was interpreted eliminates: reading a genotype-overlap percentage with no scale - 97% and 78% are host and donor only because the same-individual and different-individual ranges exist · outcome: held Quotation withheld because this record is not cleared for redistribution. the routine clinical chimerism assay, PowerPlex 16 on a genetic analyser, run on the same post-transplant samples eliminates: the new method being its own witness - the chimerism call is checked against the assay currently trusted for that decision, in both patients · outcome: held Quotation withheld because this record is not cleared for redistribution. the remaining fresh PBMCs from the same donor frozen and re-run three weeks later, compared with the fresh library gene by gene and cluster by cluster eliminates: freeze-thaw as an explanation of any composition difference in the frozen repository specimens the method is meant to be used on · outcome: held Quotation withheld because this record is not cleared for redistribution. ten bead-enriched, separately profiled subpopulations of PBMCs from the same donor, each clustered on its own and checked against FACS eliminates: substructure being an artefact of clustering a mixed population - a purified sample that splits would indict the pipeline, and most of them did not split · outcome: held Quotation withheld because this record is not cleared for redistribution. Seurat, an independently written clustering package, run on the same 68k gene-barcode matrix eliminates: the cell-type structure being specific to the authors' own PCA and k-means choices - a second package agreeing on some boundaries and not others locates which ones the data supports · outcome: partial Quotation withheld because this record is not cleared for redistribution. GEM GEM Quotation withheld because this record is not cleared for redistribution. UMI UMI Quotation withheld because this record is not cleared for redistribution. multiplet rate multiplet rate Quotation withheld because this record is not cleared for redistribution. multiplets multiplets Quotation withheld because this record is not cleared for redistribution. cell capture rate cell capture rate Quotation withheld because this record is not cleared for redistribution. cross-talk cross-talk Quotation withheld because this record is not cleared for redistribution. technical variation technical variation Quotation withheld because this record is not cleared for redistribution. biological variation biological variation Quotation withheld because this record is not cleared for redistribution. ERCC ERCC Quotation withheld because this record is not cleared for redistribution. SNVs SNVs Quotation withheld because this record is not cleared for redistribution. chimerism chimerism Quotation withheld because this record is not cleared for redistribution. genotype overlap genotype overlap Quotation withheld because this record is not cleared for redistribution. Seurat Seurat Quotation withheld because this record is not cleared for redistribution. Poisson distribution Poisson distribution Quotation withheld because this record is not cleared for redistribution. multiplet rate - AS MEASURED BY SPECIES MIXING: twice the share of barcodes carrying significant UMI counts from both genomes, which can only see cross-species doublets Quotation withheld because this record is not cleared for redistribution. multiplet rate - AS MEASURED BY SNV: twice the share of cells whose SNV counts fall between the two cell lines, which still cannot see a doublet of two like cells Quotation withheld because this record is not cleared for redistribution. the method's level of detection - AS STATED IN RESULTS: 3%, with minor genotypes identified at >95% sensitivity and PPV and nothing recovered below it Quotation withheld because this record is not cleared for redistribution. the method's level of detection - AS STATED IN METHODS: 4% of 6,000 cells, from the same in silico mixing experiment; the paper never reconciles the two Quotation withheld because this record is not cleared for redistribution. variation in UMI counts - IN A DROPLET HOLDING A CELL: biological and technical together, CV2 ~11-19% Quotation withheld because this record is not cleared for redistribution. variation in UMI counts - IN A DROPLET HOLDING SYNTHETIC RNA: technical only, by construction, CV2 ~7% Quotation withheld because this record is not cleared for redistribution. critiques → existing scRNA-seq methods - microfluidic approaches with limited throughput, plate methods needing FACS into many plates, and droplet methods requiring custom devices and reagents Quotation withheld because this record is not cleared for redistribution. critiques → clinical chimerism testing by PCR on flow-sorted populations, which is bounded by cell-surface markers, needs sufficient cell counts per population, and is not intended to detect minimal residual disease Quotation withheld because this record is not cleared for redistribution. extends → genotype-based cell assignment, by inferring both genomes de novo from the transcriptome rather than requiring DNA sequencing or large copy-number changes known in advance Quotation withheld because this record is not cleared for redistribution. confirms → Klein et al.'s estimate that technical variance accounts for about half of the total variance in droplet scRNA-seq Quotation withheld because this record is not cleared for redistribution. qualifies → its own PBMC classification - the correlation-based assignment blurs T-cell subpopulation boundaries, and an independent package fails on a different pair Quotation withheld because this record is not cleared for redistribution. technical variation —distinct-from→ biological variation technical variation biological variation differs in: whether the source of the variance is the cell or the instrument · the relation itself is separately evidenced · strength of the relation: 0.95 · differs in (stated by the tagger, no quotation): whether the source of the variance is the cell or the instrument · the relation stated: an ERCC droplet contains no cell, so nothing biological can vary in it Quotation withheld because this record is not cleared for redistribution. multiplet rate —operationalised-as→ UMI multiplet rate UMI the relation itself is separately evidenced · strength of the relation: 0.9 · the relation stated: defined as twice the rate of barcodes carrying significant UMI counts from both genomes Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span defines the multiplet rate FROM UMI counts; operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. multiplets —contrasted-with→ Poisson distribution multiplets Poisson distribution differs in: whether cells enter droplets independently, or arrive already stuck together · the relation itself is separately evidenced · strength of the relation: 0.9 · differs in (stated by the tagger, no quotation): whether cells enter droplets independently, or arrive already stuck together · the relation stated: the imaged rate of multiple encapsulations exceeded the Poisson expectation by 16% Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the multiplet rate exceeded the expected value; contrasted-with 'Poisson distribution' is loose. Quotation withheld because this record is not cleared for redistribution. Genotype overlap averages about 98% between samples of the same individual and about 73% between different individuals. Genotype overlap averages about 98% between samples of the same individual and about 73% between different individuals. Quotation withheld because this record is not cleared for redistribution. technical variance accounts for about half of total variance, measured by loading synthetic RNA into droplets in place of cells technical variance accounts for about half of total variance, measured by loading synthetic RNA into droplets in place of cells Quotation withheld because this record is not cleared for redistribution. Detecting multiplets of identical or highly similar cell types will require more sophisticated methods. Detecting multiplets of identical or highly similar cell types will require more sophisticated methods. Quotation withheld because this record is not cleared for redistribution. imaged multiple-cell encapsulation ran 16% above the Poisson expectation, with cell clumps seen during the manual count imaged multiple-cell encapsulation ran 16% above the Poisson expectation, with cell clumps seen during the manual count Quotation withheld because this record is not cleared for redistribution. marker-based and SNV-based cell assignment agree at 99% on the 293T/Jurkat mixture marker-based and SNV-based cell assignment agree at 99% on the 293T/Jurkat mixture Quotation withheld because this record is not cleared for redistribution. The SNV result was consistent with a clinical chimerism assay showing only 12% donor in the post-transplant sample. The SNV result was consistent with a clinical chimerism assay showing only 12% donor in the post-transplant sample. Quotation withheld because this record is not cleared for redistribution. twice the rate of cell barcodes with significant UMI counts from both mouse and human, where the top 1% of UMI counts counts as significant multiplet rate Quotation withheld because this record is not cleared for redistribution. the ratio of the number of cells detected by sequencing to the number of cells loaded cell capture rate Quotation withheld because this record is not cleared for redistribution. Gel bead in EMulsion - one gel bead carrying barcoded oligonucleotides, encapsulated with cells and reagents in an 8-channel microfluidic chip GEM Quotation withheld because this record is not cleared for redistribution. the inferred multiplet rate at 1,012 cell-containing GEMs in the species-mixing run 1.6% Quotation withheld because this record is not cleared for redistribution. the mean fraction of UMI counts from the other species - the barcode cross-talk level 0.9% Quotation withheld because this record is not cleared for redistribution. the error in the cell counts that feeds into every capture-efficiency figure 15–20% Quotation withheld because this record is not cleared for redistribution. genotype overlap between samples from the same individual - the positive control that makes 97% readable as host 98±0.3% value source note: THE DESCRIPTION DRAWS ON THE PAPER BEYOND THIS QUOTE: the 97% is the transplant result the control is used to read ('the post-transplant sample was 97% similar…'). Quotation withheld because this record is not cleared for redistribution. genotype overlap between different individuals - the negative control that makes 78% readable as donor 73±2% value source note: THE DESCRIPTION DRAWS ON THE PAPER BEYOND THIS QUOTE: the 78% is the transplant result the control is used to read ('The sample only shared 78% similarity with AML035 host cells'). Quotation withheld because this record is not cleared for redistribution. sensitivity sensitivity Quotation withheld because this record is not cleared for redistribution. PPV PPV Quotation withheld because this record is not cleared for redistribution. Poisson distribution Poisson distribution Quotation withheld because this record is not cleared for redistribution. multiplet rate multiplet rate Quotation withheld because this record is not cleared for redistribution. genotype overlap genotype overlap Quotation withheld because this record is not cleared for redistribution. technical variation technical variation Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1038/ncomms14049.