TY - JOUR
T1 - Preliminary evaluation of ShallowHRD performance compared to HRDetect in familial breast cancer tumors
AU - Adel Jensen, Louise
AU - Baekgaard, Caroline Hey
AU - Larsen, Mie Bohnensack
AU - Boonen, Susanne Eriksen
AU - Bak Jylling, Anne Marie
AU - Hikmat, Zainab
AU - Hao, Qin
AU - van Overeem Hansen, Thomas
AU - Pedersen, Inge Søkilde
AU - Larsen, Martin Jakob
AU - Thomassen, Mads
N1 - Publisher Copyright:
© The Author(s) 2025.
PY - 2025/8/11
Y1 - 2025/8/11
N2 - Determining the Homologous Recombination Deficiency (HRD)-status of a malignant tumor is central in predicting patient response to specific treatments. Therefore, precise and cost-effective tools are needed for clinical implementation. HRDetect is widely regarded as a golden standard for determining HRD-status. In contrast, ShallowHRD is a simpler algorithm. However, it offers a more economical alternative optimized for Formalin-Fixed, Paraffin-Embedded tissue (FFPE) and potentially useful for most breast cancer patients. Data from shallow whole-genome sequencing (1-5X) on FFPE tissue and whole-genome sequencing (50X, and additionally downscaled to 5X) on fresh frozen tissue from 19 patients were analyzed using ShallowHRD and compared to the HRD-status attained by HRDetect using Receiver Operating Characteristic (ROC) curve analysis. Further, Spearman rank correlation was calculated to estimate the correlation between ShallowHRD and HRDetect scores, as well as between the three ShallowHRD datasets. The comparison of ShallowHRD to HRDetect displayed a significant specificity (85.7–100%) and sensitivity (80%) in all data groups. The ROC curve analyses illustrated that ShallowHRD displayed an area under the curve statistically similar to what was previously reported for HRDetect in all three data sets. The Spearman correlation also indicated significant correlation between ShallowHRD and HRDetect scores for the three datasets (HRDetect vs. FFPE (1-5X) (ρ = 0.68, p = 0.0013), Fresh Frozen (5X) (ρ = 0.58, p = 0.0086), and Fresh Frozen (50X) (ρ = 0.50, p = 0.029)). The ShallowHRD analysis was of good quality in all data groups, and the ShallowHRD scores were similar across data groups. One sample was incorrectly labeled as HRD-negative by ShallowHRD, but it contained one variant of unknown significance (VUS) in RAD51D, requiring further investigation. HRD-status from ShallowHRD correlated well with HRDetect output in this preliminary study, potentially making ShallowHRD an accurate, efficient, and more economical alternative for clinical use. However, further examination and validation in larger cohorts are required.
AB - Determining the Homologous Recombination Deficiency (HRD)-status of a malignant tumor is central in predicting patient response to specific treatments. Therefore, precise and cost-effective tools are needed for clinical implementation. HRDetect is widely regarded as a golden standard for determining HRD-status. In contrast, ShallowHRD is a simpler algorithm. However, it offers a more economical alternative optimized for Formalin-Fixed, Paraffin-Embedded tissue (FFPE) and potentially useful for most breast cancer patients. Data from shallow whole-genome sequencing (1-5X) on FFPE tissue and whole-genome sequencing (50X, and additionally downscaled to 5X) on fresh frozen tissue from 19 patients were analyzed using ShallowHRD and compared to the HRD-status attained by HRDetect using Receiver Operating Characteristic (ROC) curve analysis. Further, Spearman rank correlation was calculated to estimate the correlation between ShallowHRD and HRDetect scores, as well as between the three ShallowHRD datasets. The comparison of ShallowHRD to HRDetect displayed a significant specificity (85.7–100%) and sensitivity (80%) in all data groups. The ROC curve analyses illustrated that ShallowHRD displayed an area under the curve statistically similar to what was previously reported for HRDetect in all three data sets. The Spearman correlation also indicated significant correlation between ShallowHRD and HRDetect scores for the three datasets (HRDetect vs. FFPE (1-5X) (ρ = 0.68, p = 0.0013), Fresh Frozen (5X) (ρ = 0.58, p = 0.0086), and Fresh Frozen (50X) (ρ = 0.50, p = 0.029)). The ShallowHRD analysis was of good quality in all data groups, and the ShallowHRD scores were similar across data groups. One sample was incorrectly labeled as HRD-negative by ShallowHRD, but it contained one variant of unknown significance (VUS) in RAD51D, requiring further investigation. HRD-status from ShallowHRD correlated well with HRDetect output in this preliminary study, potentially making ShallowHRD an accurate, efficient, and more economical alternative for clinical use. However, further examination and validation in larger cohorts are required.
KW - Breast cancer
KW - Cancer genomics
KW - Homologous recombination deficiency (HRD)
KW - HRDetect
KW - ShallowHRD
KW - Whole genome sequencing (WGS)
KW - Homologous Recombination/genetics
KW - Humans
KW - Middle Aged
KW - Algorithms
KW - Female
KW - ROC Curve
KW - Adult
KW - Breast Neoplasms/genetics
KW - Whole Genome Sequencing/methods
UR - https://www.scopus.com/pages/publications/105012927912
U2 - 10.1038/s41598-025-14122-9
DO - 10.1038/s41598-025-14122-9
M3 - Journal article
C2 - 40790321
AN - SCOPUS:105012927912
SN - 2045-2322
VL - 15
SP - 29442
JO - Scientific Reports
JF - Scientific Reports
IS - 1
M1 - 29442
ER -