Long-Read Sequencing Outperforms Short-Read Sequencing in Detecting Most Structural Variations

Abstract

Structural variations (SV) are common in the cancer genome and play critical roles in regulating tumorigenesis. In the past decades, many SVs have been detected through analyses of whole-genome sequencing (WGS) data generated mainly by Illumina paired-end short-read sequencing (SRS). Recent advances in long-read sequencing (LRS) techniques provide exciting opportunities for SV detection. However, a comprehensive analysis of the pros and cons of LRS and SRS in detecting SVs in a cancer genome is still lacking. Here, we performed WGS of the LNCaP prostate cancer cell line through LRS using the Oxford Nanopore Technology and called main SVs, which were compared to those derived from publicly available LNCaP SRS data. Strikingly, LRS is superior in detecting insertions of all sizes and deletions of <1000 bp long, whereas SRS is very useful in capturing long deletions, taking advantage of its paired-end reads. LRS identified more precise breakpoints of detected SVs. In addition, we found that SRS called many duplications and inversions, most of which were not confirmed by LRS, likely due to ambiguity in SRS read alignment to repetitive regions, leading to errors in SV calling. In conclusion, LRS outperformed SRS in detecting most SVs, except deletions longer than LRS read lengths. Our study highlights the advantages of LRS in resolving complex genomic rearrangements and underscores its potential for improving SV detection in cancer genomics.