An AI model created by researchers at King’s can detect signs of faulty DNA-repair mechanisms in ovarian cancer cells, potentially enabling a low-cost, widely accessible diagnostic tool.
By analysing tissue images, the model recognised tumours with defective DNA-repair mechanisms, which could help clinicians select treatments that exploit this vulnerability.
We are unlocking information hidden in microscopic images of cancer tissue that could transform how we match patients to treatments. By using AI to reveal defects in DNA repair, we could make precision cancer medicine faster, more affordable and accessible to more women with ovarian cancer, not just in the UK but around the world.
Dr Heba Sailem, Reader in Computational Pathology at King’s College London
The study focused on high-grade serous tubo-ovarian carcinoma (HGSOC). Homologous recombination deficiency (HRD), often arising from BRCA1/BRCA2 or related mutations, impairs DNA-damage repair and increases vulnerability to platinum chemotherapy and PARP inhibitors. HRD is therefore a useful biomarker for treatment selection, but current tests require genomic analysis that is costly and less accessible in resource-limited settings.
In the study, published in Nature Communications and led by Dr Heba Sailem, researchers developed HRDPath. Rather than sequencing DNA, HRDPath analyses high-resolution digital histopathology images. It was trained on images from patients with known genomic HRD status and correctly identified approximately 91% of HRD-positive cases, particularly when multiple images per patient were considered.
HRD-positive tumours showed subtle morphological differences, including rounder nuclei and denser cellular packing in certain regions. Defects in DNA repair therefore appear linked to tissue-architecture changes recognisable by AI but difficult to detect by conventional microscopy.
If validated in larger cohorts, HRDPath could use images already acquired for diagnosis, reducing reliance on expensive genomic testing and widening access to platinum and PARP-inhibitor selection for patients whose tumours are likely to benefit.
Source: Kings College London
