Fig. 1

Overview of the SQUID algorithm. Based on the alignments of RNA-seq reads to the reference genome, SQUID partitions the genome into segments (step 1), connects the endpoints of the segments to indicate the actual adjacency in transcript (step 2), and finally reorders the endpoints along the most reliable path (step 3). Thick black lines are genome sequences or segments. Grey, red, and cyan short lines are read alignments, where grey represents concordant alignment, and red and cyan represent discordant alignments of different candidate TSVs. Vertical dashed lines are the separation boundaries between genome segments, and the boundaries are derived based on read alignments. The heads of genome segments are denoted by As, Bs, etc., and the tails are denoted by At, Bt, etc. Step 2: Each read alignment generates one edge between segment endpoints. An edge is added in the following way. When traversing the genome segments along the edge to generate a new sequence, the read can be aligned concordantly onto the new sequence. Multi-edges are collapsed into one weighted edge, where the weight is the number of reads supporting that edge. Red and cyan edges correspond to different candidate TSVs. Step 3: Genome segments are reordered and reoriented to maximize the total number of concordant alignments (concordant edge weights) with respect to the new sequence. Step 4: Discordant edges that are concordant after rearrangement are output as TSVs (in this case, both red edges and cyan edges are output). chr chromosome, TSV transcriptomic structural variant