Comparing HRD Signals in TNBC
In this cohort, HRD classifiers were broadly concordant, yet differences in input modality, preprocessing, and training context materially shaped individual classifications, supporting careful workflow and threshold standardization rather than treating methods as interchangeable.
> Research explainer: This briefing examines verified primary research published 62 days before the briefing date. It is not a same-day research update and does not provide medical advice.
Evidence
This research explainer examines a study that compared seven homologous recombination deficiency (HRD) classification methods in 235 patients with early-stage, treatment-naive triple-negative breast cancer (TNBC) from a reported population-based cohort in southern Sweden. The available material included RNA sequencing, whole-genome sequencing, RAD51 foci staining on tissue microarrays, and digital H&E whole-slide images. The seven methods covered sequencing-based, copy-number-based, functional, mRNA-based, and image-based approaches. (pmid:42351273)
The comparison found substantial overall concordance across the methods, but it also identified method-specific discordance. HRDetect and CHORD, the sequencing-based methods, and scarHRD showed the greatest classification agreement. This result matters because the study did not portray all HRD outputs as interchangeable at the individual-tumor level, even where overall agreement was substantial. (pmid:42351273)
The study linked part of the disagreement among sequencing-based and copy-number-based assessments to inadequate tumor cell content, sequencing depth, and data-processing steps such as segmentation. For mRNA- and image-based classifications, discordance appeared associated with molecular subtype features. The authors described this as evidence that training-cohort context can incorporate signals, including mRNA-expression patterns, that are not specific to HRD status. (pmid:42351273)
All seven methods had approximately similar prognostic performance for invasive disease-free survival in the subset treated with adjuvant standard-of-care chemotherapy. The analysis used prognosis as its endpoint; it did not establish treatment benefit, clinical utility, or a preferred method for selecting treatment. (pmid:42351273)
An exploratory comparison also evaluated 18 selected tumors with the FDA-approved Myriad myChoice CDx assay. This small selected subset provides a limited cross-assay reference within the study rather than a broad validation of every method against that assay. (pmid:42351273)
Analysis — Measuring Signal Across Platforms
The central lesson is about measurement rather than a single winner among classifiers. In this cohort, agreement was strongest among HRDetect, CHORD, and scarHRD, while technical inputs and analytical choices helped explain some disagreements. That pattern supports a bounded interpretation: an HRD label can depend partly on the evidence a platform receives and the workflow used to convert that evidence into a classification. (pmid:42351273)
The distinction between overall concordance and individual discordance is especially important. Broad agreement can coexist with consequential disagreement for particular samples when tumor cell content, sequencing depth, segmentation, or platform-specific training signals differ. The mRNA- and image-based findings further indicate that molecular subtype features may influence classifications in ways not specific to HRD. (pmid:42351273)
Accordingly, the study supports optimizing data-processing workflows and threshold definitions to improve consistency, comparability, and reproducibility across HRD platforms in this setting. Its approximately similar prognostic performance finding should be read narrowly: it concerns invasive disease-free survival in the adjuvant-chemotherapy subset and does not resolve clinical utility or treatment-selection questions. (pmid:42351273)
Limitations
The cohort was limited to 235 patients with early-stage TNBC from southern Sweden, so this study does not establish generalizability to other populations, cancer stages, or cancer types. Only 150 of 235 samples could be classified by all seven methods. (pmid:42351273)
The Myriad myChoice CDx comparison involved only 18 selected tumors, limiting the scope of that exploratory assessment. In addition, the supplied study material does not provide complete method-specific performance estimates, confidence intervals, or detailed subgroup results. (pmid:42351273)
Finally, the prognostic analysis was confined to patients treated with adjuvant standard-of-care chemotherapy. It does not establish treatment benefit or clinical utility for any HRD method, and the study’s findings should not be used for patient-specific conclusions. (pmid:42351273)
Evidence boundary
This one-source briefing is limited to what the cited study reports. It does not establish independent confirmation, broader clinical effectiveness, or patient-specific guidance. The design, population, measurements, and follow-up described in that source define the evidence boundary. This summary provides research context and is not medical advice. The evidence should be read as a bounded report of the study rather than as a conclusion about other populations, settings, interventions, or outcomes. Any possible connection to disease mechanisms remains limited to the measurements and interpretations documented by the cited authors. Terms describing associations, responses, or biological patterns retain the meaning and uncertainty given in that source.
No inference beyond the cited source is made here.