Rationale and overview of analysis approach. (A) Cartoon of a hypothetical locus encoding alternatively spliced transcripts, illustrating how junction spanning reads map unambiguously to specific introns. Read 1 could have originated from the 2nd exon of isoform A or B, or the intron of isoform C; while read 2 could only have originated from isoform A and the indicated splice junction. (B-E) Flowcharts of analysis steps performed in Spanki. Input data listed at the top, format in parentheses, and calls to external programs indicated (bold). (B) Flowchart of simulation methods. A two step process begins with modeling error profiles based on a permissive Bowtie  alignment. These error models are used by the simulator to generate reads. (C-E) Flowcharts of quantification and comparison methods. The first step is junction quantification (C), where alignments are performed, junction alignments are curated, and junction coverages are calculated. Splicing event quantification (D), where a set of transcript models (from annotation or computed using a program such as Cufflinks ), are used to characterize pairwise splicing differences (“splicing events”). These events are merged with junction coverage data to quantify the mutually exclusive paths defined for each event. Splicing event comparison (E) uses these tabulated event-level quantifications to compare between replicates, and between pooled results for each sample, by Fisher’s Exact Test on inclusion and exclusion junction counts.
Sturgill et al. BMC Bioinformatics 2013 14:320 doi:10.1186/1471-2105-14-320