Skip to main content
Figure 3 | BMC Bioinformatics

Figure 3

From: SpliceJumper: a classification-based approach for calling splicing junctions from RNA-seq data

Figure 3

Illustration of paired-end reads that are used to guide the alignment and filter out false splicing junctions. A, B, C and D are four called out splicing sites. 1(1a and 1b), 2(2a and 2b), and 3(3a and 3b) are three pairs. 1a, 2a, and 3b are fully mapped, while 1b, 2b, and 3a are split-mapped. The right end of 1a is aligned at site E, and [E,D] forms a focal region to align the clipped segment of 1b. Insert size between 1a and 1b can be used to guide the alignment of the clipped segment of 1b: the clipped segment is short and can be aligned at both position A and C, but the alignment at C is wrong because otherwise 1a and 1b form a discordant pair which has low probability. Thus alignment at position A is chosen. Paired-end reads can also be used to filter out false splicing junctions. Splicing junction [C,D] is considered as a false one, because although read 2b is split-mapped at position C and D, 2a and 2b will form a discordant pair if the alignment of 2b is correct. Thus a conflict happens, and SpliceJumper will report [C,D] a false positive. In contrast, splicing junction [A,B] is considered as a true splicing junction.

Back to article page