Figure 2

The asMIP data correctly characterize splicing controls and correlate with qPCR measurements. Receiver Operating Characteristic (ROC) curves comparing asMIP data with (A) junctions known from the literature to be alternatively spliced and (B) qPCR data. Data from asMIPs quantified using sequencing is black, asMIP data from arrays is blue; R-score data are dashed lines; M-score data are solid lines. (A) An analysis of 157 tissue specific splicing events: 58 positive alternative splicing controls [10, 20] and 99 predicted constitutive negative controls [21]. For this plot, only relevant pairs of tissues, known to show alternative splicing for the genes assayed, were included in the calculation of R- and M-scores; skeletal and stomach data were used for tropomyosin genes, stomach and frontal lobe for brain specific genes, placenta and frontal lobe for ERC1. Absolute values of these scores were used. To generate the standard for comparison, positive spicing controls were assigned a value of 1, negative constitutive controls were 0. (B) An analysis of the 235 splicing events that were studied using both qPCR and asMIPs. For this plot we transformed ternary splice scores (1, 0, and -1) into binary scores (non-zero versus zero). Because there were cases where the qPCR and asMIP scores had the opposite sign, the ROC curves do not reach the upper right-hand corner of the graph. Individual ROC curves for positive or negative splice calls are located in Additional File 4. The solid black circle marks the point on the M-score sequencing line that corresponds to the +/-1.3 cutoff, which was used to identify alternative splicing.