Fig. 2: Evaluation of computational RNA counting strategies using molecular spikes.
From: Molecular spikes: a gold standard for single-cell RNA counting
a–d, Counting difference between number of unique spike identifiers and quantified 10x Genomics UMIs at varying mean expression levels. Colored lines indicate the mean (n = 100 in silico cells) counting difference per UMI length shaded by the standard deviation. Counting difference is expressed in absolute numbers (a,c) or as a percentage of the mean spUMI count (b,d), and UMI counts were computed without error-correction (a,b) or corrected in adjacency mode (hamming distance 1) (c,d). e,f, Comparison of edit distance (hamming distance) for adjacency error correction of UMIs of length 8 or 10 nt. Lines indicate the mean (n = 100 in silico cells) difference in quantification between spUMIs and UMIs shaded by the standard deviation in absolute scale (e) or relative to the mean (f). g,h, Evaluation of computational UMI collapse methods adjacency, adjacency-singleton, adjacency-direction and cluster at edit distance 1 and UMI lengths of 8 or 10 nt. Lines indicate the mean (n = 100 in silico cells) difference in quantification between spUMIs and UMIs shaded by the standard deviation in absolute scale (g) or as a percentage relative to the mean (h).
