PMC

(a – d): FMO controls showing the increase in MFI of Podoplanin (a), CD31 (b), Lyve-1 (c), and PDL-1 (d) of doublets relative to singlets due to background is negligible relative to actual protein expression. n = 4 mice per group; data are represented as mean ± standard error of the mean. For Podoplanin MFI (a), singlet FMO vs. doublet FMO, p > 0.9999, singlet FMO vs. EAE singlets, p = 0.0055, singlet FMO vs. EAE doublets, p < 0.0001, doublet FMO vs. EAE singlets, p = 0.0057, doublet FMO vs. EAE doublets, p < 0.0001, EAE singlets vs. EAE doublets, p = 0.0016; for CD31 MFI (b), singlet FMO vs. doublet FMO, p = 0.0130, singlet FMO vs. EAE singlets, p < 0.0001, singlet FMO vs. EAE doublets, p < 0.0001, doublet FMO vs. EAE singlets, p < 0.0001, doublet FMO vs. EAE doublets, p < 0.0001, EAE singlets vs. EAE doublets, p < 0.0001; for Lyve-1 MFI (c), singlet FMO vs. doublet FMO, p = 0.0182, singlet FMO vs. EAE singlets, p < 0.0001, singlet FMO vs. EAE doublets, p < 0.0001, doublet FMO vs. EAE singlets, p < 0.0001, doublet FMO vs. EAE doublets, p < 0.0001, EAE singlets vs. EAE doublets, p = 0.7736; for PD-L1 MFI (d), singlet FMO vs. doublet FMO, p = 0.9828, singlet FMO vs. EAE singlets, p = 0.0003, singlet FMO vs. EAE doublets, p < 0.0001, doublet FMO vs. EAE singlets, p = 0.0005, doublet FMO vs. EAE doublets, p < 0.0001, EAE singlets vs. EAE doublets, p < 0.0001; one-way ANOVA with Tukey’s post-hoc multiple comparisons test.

A, B: Doublet enrichment in a toy example. A: Proportion of different doublet types from random expectations based on the cell type abundances. B: The fold-enrichment over this expectation in two different doublet enrichment scenarios. C, D: Performance of the cluster stickiness tests (C) and tests for enrichment of specific combinations (D) using different underlying distributions.

(A) Schematic overview of genotyped cell mixing experiment.
(B) Force-directed graph layout of the profiled cells. Black points indicate ground truth doublets identified by Demuxlet as barcodes associated with polymorphisms from more than one individual ().
(C) Force-directed graph layout of ground truth doublet score, defined as the fraction of a cell’s neighbors that are mixed genotyped doublets.
(D) Application of Scrublet to the transcriptomic data. After calculating doublet scores (i), the histogram of scores for simulated doublets was used to determine a threshold for detection of neotypic doublets (ii). Applying this threshold to observed cell barcodes (iii) yielded doublet predictions for each transcriptome (iv). ϕ_D, predicted detectable doublet rate. See also .
(E) Comparison of Scrublet to the ground truth doublet score, colored by genotype-based doublet labels (singlets, gray; doublets, black).
(F) Comparison of detectable doublet fraction (solid black line) and actual recall (dashed black line) for a range of doublet scorethresholds and the corresponding precision (red line). TP, true positive; FN, false negative; FP, false positive.
(G) Alternative doublet prediction based on co-expression of marker genes of distinct cell types. Upper: force-directed graph layout with cells colored by marker overlap score. Lower: histograms of marker overlap score for ground truth singlets (gray) and doublets (red).
(H) Alternative doublet prediction based total transcript counts. Upper: force-directed graph layout with cells colored by total counts. Lower: histograms of total counts for ground truth singlets (gray) and doublets (red).
(I) ROC curves (upper) and AUC scores (lower) for various doublet prediction methods. “S+TC” and “S+Local TC” are linear combinations of the Scrublet score and total counts or the Scrublet score and total counts relative to neighboring cells, respectively (see for details).

Conditional hazard analyses. (A–F) Analysis of 18 mitral cells, recorded from male rats, which fired with frequent doublet spikes. (A) Shows the mean (SE) hazard function; the function is bimodal, reflecting the frequent doublet firing. (B–F) Show mean conditional hazards, (B) shows the conditional hazards reflecting excitability after either a sequence of two long intervals (LL) or a long interval followed by two short intervals (long‐doublet‐doublet, LDD). Excitability is reduced for about 50 msec after LDD compared to after LL. (C) A doublet is more likely to follow a sequence of long‐doublet‐long (LDL) than two long intervals (LL). (D) A doublet is more likely to follow a sequence of doublet‐long‐doublet (DLD) than two long intervals (LL). (E) A doublet is equally likely to follow a sequence of DLD as it is to follow a sequence of long‐long‐doublet (LLD). (F) A doublet is more likely to follow a sequence of LLD than LDD. (G) For these mitral cells, we calculated conditional hazard functions to show how excitability evolves after (a) two long intervals (each < 10 msec; LL), (b) a sequence of doublet, long interval, doublet (DLD), and (c) after a sequence of long interval, doublet, doublet (LDD). The solid circles show the mean (±SE) difference between the hazard after LL and that after DLD, and the open circles the mean difference between the hazard after LL and that after LDD. After the first doublet of a train (i.e., after DLD) there is an enhanced chance of another spike occurring within 10 msec; but if this does not occur, then there is a reduced excitability for about 25 msec, consistent with the interpretation that a doublet spike triggers a fast DAP and a slow AHP. After two doublets (i.e., after LDD) there is generally a lower chance of another spike occurring within 10 msec, and there seems to be no prolongation of the AHP. (H) Examples of sequences LDL and LDD. 
Data File (Figure S5) Excel workbook giving, for 18 doublet cells recorded from male rats, the full hazard functions (sheet labeled “Full haz”); and the corresponding conditional hazards after DLD, LLD, LL, LDL, LDD, and DDD in the sheets with these names, and (on “DLD vs. LL”) the differences between hazards after DLD and LL.