Resolving structural variations missed by short-read sequencing uncovers their pathogenicity.

Short-read genome sequencing (sr-GS) affords efficient and accurate characterisation of apparently balanced chromosomal rearrangement (ABCR) breakpoints except in 9%-11% of cases that remain undetectable. Among 117 ABCR that we studied in patients with abnormal phenotype, 14 (11.9%) could not be detected by our current strategy including sr-GS, alignment against the GRCh38 reference genome and structural variant (SV) detection using Breakdancer V.1.4.5. These were all reciprocal translocations, 10 of which implicated constitutive heterochromatin, acrocentric short arms or pericentromeric regions. We re-aligned the sequencing data against the T2T-CHM13 V.2.0 reference genome and re-analysed them using five other SV callers (DELLY, GRIDSS, LUMPY, Manta and SvABA). In addition, 11 ABCRs were further characterised using FISH, linked-read sequencing, long-read sequencing or optical genome mapping, either isolated or combined. We were able to characterise the breakpoints at the bp level for 12 translocations and identify specific breakpoint patterns using Integrative Genome Viewer (IGV). In each translocation, at least one breakpoint involved highly repetitive elements such as alpha-satellites, segmental duplications, satellite repeats or other poorly mapped regions. For six out of 12 patients, one of the breakpoints could explain the phenotype either by gene disruption (CAMTA1, DYRK1A, NLGN4X) or position effect (BMP2, DIAPH2, SIX3). Failure of sr-GS is due to highly repetitive genomic regions at SV breakpoints, either absent from the reference genome or not attributed to a unique position. The resolution of ABCRs is essential to patients' care since it allowed us to conclude to a pathogenic variant in 50% of patients.