Fig. 5

Schematic representation of two pair-wise correlation measures that have been used to derive co-splicing networks (isoform correlations and cosplicing correlations). In the top track, exon count data for two genes is used to infer the isoform abundance for each subject (shown in colors). Gene a and gene b contain 4 and 3 exons, under 2 and 3 isoforms models, respectively. As, isoform ratios between genes a and b are obtained, all possible correlations (2*3 = 6) among the isoforms of both genes can be computed. The example illustrates a significant correlation between isform 2 of gene a and isoform 3 of gene b. In this work, we take the lower track to compute the gene correlations of the co-splicing networks. Here, exon count data is directly used to determine differences between individuals at each gene. No inference of specific isoform models is required and all data is used to determine the distance matrix of individuals. The matrix-based correlations test weather the similarities in the exon count distributions across subjects are kept between genes. The figure illustrates a likely high co-splicing correlation between the genes derived from the fact that, in both genes, the curve of the exon count distributions for subject 3 (red) is substantially different to the curves for subjects 1 and 2 (green and blue). With one single measure the co-splicing correlation between genes, based on distance matrix, informs on the possible coregulation of splicing across all isoforms