Fig. 1

Single-cell splicing and considered features for modeling splicing rates. a Two canonical splicing models. The “cell model” assumes that splicing variation is due to the differential splicing between cells, with each cell expressing one of two splice isoforms. The “gene model” corresponds to the assumption that both splice isoforms can be expressed in the same cells. b Mean-variance relationships of splicing rates in iPS cells. Shown is the standard deviation of splicing rates across cells for the same cassette exon (standard deviation of PSI) as a function of the average inclusion rate of the cassette exons across cells, considering 84 iPS cells. Solid lines correspond to the expected relationship when either assuming a “cell model” (black line) or when assuming the “gene model” (red line). c Illustration of the considered features and genomics contexts for predicting splicing variation. “A” denotes the alternative exon; “I1” and “I2” correspond to the upstream and downstream flanking introns, respectively; and “C1” and “C2” to the upstream and downstream flanking exons, respectively. The 5′ and 3′ ends (300 bp) of the flanking introns are considered separately