PMC

Percentage of bases in the targeted breakpoint hotspots for ALK and KMT2A covered at levels ranging from 5× to 1000× for all positive controls. The box and whisker plots represent the median and upper and lower quartiles for the 13 rearranged cases.

Sensitivity of Breakdancer (black), Hydra (blue), and ClusterFAST (red) to detect the breakpoints in the 13 ALK and KMT2A rearranged cases in randomly down-sampled binary sequence alignment files. Squares indicate the mean (over three random samples) sensitivity per tool, and error bars indicate SE in the mean.

Coverage profiles within the targeted breakpoint hotspots for ALK and KMT2A. The interquartile range of coverage depth at each position (blue-gray), percentage of GC content (black), and alignability (CRG 50; gray) over the targeted capture region (exons as dark blue boxes) are shown. Breakpoints located in the set of positive controls are indicated with vertical red lines. CRG, Centre for Genomic Regulation.

Effect of duplicate reads. A: The Circos plot shows all translocation events within the 151 targeted genes on the panel in the six FISH-positive KMT2A rearranged cases identified by Breakdancer (red lines) and ClusterFAST (blue) when duplicate reads were not removed before analysis. B: Box and whisker plot of log₁₀ counts of all structural variants, interchromosomal rearrangements, and ALK/KMT2A interchromsomal rearrangements, detected by Breakdancer (red) and ClusterFAST (blue) for all reads (dark red/blue) and with duplicates removed (light red/blue).

A comparison of bioinformatic tools for translocation detection. A: Discordant read pair mapping is the simplest approach for identification of translocations in NGS data and is the main method used by Breakdancer and Hydra. Using this approach, we detected rearrangements when one end of a read pair maps to the chromosome of interest and the other maps to a different chromosome or the same chromosome at a greater than expected distance. Such methods can approximate the position of the breakpoint to within approximately 100 bp but do not generate a contig of the actual breakpoint sequence. B: Soft-clipped reads are generated during sequence alignment when the far end of the read does not match the reference sequence and is then masked from further analysis. With CREST, translocations are identified in areas where increased soft-clipped reads are identified and can be assembled into a contig spanning the translocation breakpoint. Breakpoint sequences are then aligned back to the reference genome to determine the position of the breakpoint. C: ClusterFAST uses a split single end read method to identify translocations that consists of three phases. In the first phase, discordant read pairs are identified similar to Breakdancer and Hydra. In the second phase, putative translocations are further evaluated by finding one end anchored or partially mapped reads (soft-clipped) that map in the vicinity of the translocation. These reads are then split into artificial read pairs and realigned to determine whether they can be mapped. If the reads are mapped and also span the breakpoint, a contig is produced by assembling all local reads in the area. This contig is then mapped back to the reference genome.