Fig. 3: COLO829 instance 1 - complex BFB on chr3, chr6, chr10, and chr12.

a The SV breakpoints split the local genome regions of chr3, chr6, chr10, and chr20 into 8, 4, 3, and 3 segments, respectively. b List of SVs and the segments they connect. c CIRCOS diagram of the complex BFB. d Ambigram resolved the BFB path with 98x depth and 100% purity PE sequencing data. In the first stage, six BFB cycles happen on chr3 with details shown in Supplementary Fig. 11. At the same time, chr6 undergoes four BFB cycles. In the second stage, an inter-chromosomal rearrangement inserts fragments of chr10 and chr12 into the BFB path on chr3. Another translocation connects two BFB paths on chr3 and chr6, constructing the result of the complex BFB event. e The total CN error, CN accuracy, SV recall, SV precision, and SV F1-score derived by Ambigram, LINX, and BFBFinder compared to ground truth. The total CN error is the sum of all segment copy number differences between the output and ground truth. BFBFinder is marked with “*” as BFBFinder merely accepts segment CN profiles from a single chromosome, so we fit the ground truth CN profiles of BFB paths of chr3 and chr6 separately. f Results derived by Ambigram with various sequencing protocols, depths, and purities. “-” means that the inputs of Ambigram are SVs called from one sequencing protocol (SV=PE, SV=10x, SV=PB, or SV=ONT) and ground truth CNs. “Resolved” means that inferred BFB path includes all ground truth FBIs, otherwise “Unresolved''. “Resolved by inferring the virtual FBI” signifies Ambigram resolves the BFB path by recovering the undetected FBIs in low sequencing depth and tumor purity scenarios. “Resolved by utilizing read linkage” means that Ambigram cannot resolve the BFB path with CNs and detected SVs, while it can resolve the path after incorporating the linked or long-read linkage from 10x, PB, or ONT data. BFB breakage-fusion-bridge. FBI-hh fold-back inversion with head-to-head direction. FBI-tt fold-back inversion with tail-to-tail direction. DEL deletion, DUP duplication, INS insertion, TRX translocation, SV structure variation, PE paired-end, 10x 10x linked-reads, PB PacBio, ONT Oxford Nanopore.