Genome-wide loss of heterozygosity (gLOH)

Every cell carries two copies of most of its genome, one from each parent. Loss of heterozygosity is the loss of one of those copies at a locus; genome-wide loss of heterozygosity is the percentage of the interrogated genome showing it, computed from a targeted next-generation sequencing panel. It is one of the genomic scar measures of homologous recombination deficiency: a cell that cannot repair double-strand breaks accurately falls back on error-prone mechanisms, and the accumulated result is large stretches of single-copy genome. Because it measures the consequence rather than the cause, it can be raised by faults the panel does not sequence, and can be normal in a tumour that has only recently lost repair capacity.

The threshold is set by the assay, not by biology, and it is a continuous score cut into two boxes. ARIEL2 part 1 prespecified 14 percent or more as loss-of-heterozygosity high in ovarian carcinoma; that cut point was derived and validated for one assay in one disease and does not transfer unchanged. Pan-cancer work from the same platform showed that biallelic BRCA1 and BRCA2 alterations are associated with elevated genome-wide loss of heterozygosity across many tumour types while monoallelic alterations are not (Sokol 2020), and that the association extends beyond BRCA to a core set of homologous recombination repair genes including BARD1, PALB2, FANCC, RAD51C and RAD51D, particularly in breast, ovarian, pancreatic and prostate cancer, with an independent contribution from TP53 loss (Westphalen 2022).

In prostate cancer it is the number that shows the homologous recombination repair gene list is not one biomarker. Across 3,476 clinically advanced prostate tumours profiled in routine practice, BRCA1, BRCA2, ATR and FANCA alterations were associated with high genome-wide loss of heterozygosity, whereas CDK12-altered tumours, about 6 percent of the disease, were infrequently loss-of-heterozygosity high (Chung 2019). CDK12 is on the gene list that qualifies men for PARP inhibitors in several licences, and by this measure those tumours are not homologous recombination deficient in the sense a PARP inhibitor needs. That, together with TRITON3's hazard ratio of 0.95 in the ATM subgroup against 11.2 versus 6.4 months in the BRCA subgroup, is why the gene list is increasingly read gene by gene rather than as a single qualifying category.