Automated Detection of Arm-Level Alterations for Individual Cancer Patients in the Clinical Setting.

Abstract Copy number alterations (CNAs) are genetic events that promote tumor initiation and progression and are used in clinical care as diagnostic, prognostic and predictive biomarkers. Based on the length of the alteration, they are roughly classified as "focal" and "arm-level" alterations. While genome-wide techniques to detect arm-level alterations are gaining momentum in hospital laboratories, the high precision and novelty of these techniques pose new challenges: There is no consensus on the definition of an arm-level alteration and a lack of tools to compute them for individual patients. Based on 376 clinical samples analyzed with the OncoScan FFPE assay, we observed a bimodal distribution of the percentage of bases with CNAs within a chromosomal arm, with the second peak starting at 90% of arm length. We tested two approaches for the definition of arm-level alterations: sum of altered segments (SoS) >90% or the longest segment (LS) >90%. The approaches were validated against expert annotation of 25 clinical cases. The SoS method outperformed the LS method with a higher concordance (SoS: 95.2 %, LS: 79.9 %). Some of the discordances were ultimately attributed to human error, highlighting the advantages of automation. The increase in reliability led to the development of a publicly available software and its inclusion into routine clinical practice in the Geneva University Hospitals.