Cancer genome standards for long-read sequencing using cancer cell line mixtures

Long-read sequencing (LRS) improves genome alignment and facilitates resolving variants in genomic regions of low complexity, making it a promising approach for cancer variant detection and biomarker discovery.

Here, we evaluate the performance of LRS using Oxford Nanopore sequencing to detect somatic variants across different tumour purities and sequencing depths by comparing to somatic variants from short-read sequencing. We generated experimental mixtures of cancer cell lines and matched normal cell lines to simulate 10 tumour purities (ranging from 0 to 100% tumour content). This resulted in 22 samples, which were sequenced using whole-genome LRS to a targeted depth of 60x, and somatic variants were identified. We down-sampled data to explore the optimal read depth for somatic variant detection. Our results show long-read variant calling tools achieve recall rates comparable to short-read gold standards. While tumour sequencing depth of 30x-60x is generally sufficient for detecting common variants, particularly structural variants, sequencing the matched normal sample at adequate depth is crucial for accuracy. Notably, we found variants unique to LRS that may represent real and previously undetected events.

This study highlights key factors for optimising cancer genome sequencing with LRS and provides a comprehensive dataset of cell line mixtures for future research and tool development.