Characterization of rare genomic structural variants across 2,981 genomes reveals significant involvements in recessive conditions.

Structural variants (SVs) underlie genomic diversity and human diseases; however, current short-read sequencing-based methods experience challenges in precisely identifying and determining SV composition, which often lead to incorrect interpretation and/or missing detection of clinically significant variants. Here, we introduce SCARD (split-read, chimeric read-pairs, and read-depth difference), an integrated approach optimized for the detection and characterization of rare SVs using short-read genome sequencing data. Firstly, we evaluated rare SV (with minor allele frequency <1%) detection performance using 16 clinical cases with 18 previously ascertained exonic deletions/duplications in dominant/recessive conditions. SCARD accurately detected all 18 exonic deletions/duplications, ranging from 490 bp to 24.5 kb, confirmed by gap-PCR and/or Sanger sequencing, outperforming other methods that only achieved 15/18 at best. In addition, SCARD characterized rare SVs from 2,504 unrelated individuals and 477 trios/probands from the 1000 Genomes Project, and compared them to those from 13 SV detection methods (13callers). Among the 2,504 genomes analyzed, we identified 429,259 rare simple deletions/duplications and 7,795 structural rearrangements, with 10.7% of deletions/duplications (46,032/429,259) and 40.1% of structural rearrangements (3,126/7,795) being novel to those from the 13callers. Notably, 45.2% (1,133/2,504) of the subjects carried likely loss-of-function SVs involving autosomal recessive disorders related genes. In 477 trios, the overall de novo rate for rare SVs was 0.5% (383/77,349). Finally, to evaluate the accuracy of rare SV calling, we compared our SVs in 15 cases to those of high-quality long-read sequencing datasets. The accuracies of detecting deletions/duplications, structural rearrangements, and all SVs by SCARD were 98.0% (2,858/2,917), 94% (47/50), and 97.9% (2,905/2,967), respectively. In comparison, 13callers missed or misinterpreted 325 of 2,905 validated rare SVs (11.2%) reported by SCARD. Overall, our study demonstrates that SCARD enables comprehensive, genome-wide identification and precise delineation of rare SVs, particularly for those affecting exons, to reveal a high carrier rate of autosomal recessive disorders contributed by rare SVs.