L26/P-659 Haplotype-aware detection of haploidy and triploidy from low-coverage WGS-based PGT-A data

Abstract

Abstract Study question Can haplotype-based methods detect genome-wide ploidy abnormalities (haploidy and triploidy) that are missed by conventional coverage-based analysis of low-coverage whole-genome sequencing (WGS)-based PGT-A data? Summary answer Our method (LD-PGTA) enables accurate detection of hidden signatures of haploidy and triploidy from ∼0.03× WGS data, achieving high specificity with moderate sensitivity. What is known already Conventional coverage analysis of WGS-based PGT-A data cannot detect uniform genome-wide ploidy errors such as haploidy, genome-wide uniparental isodisomy, or triploidy because global changes in copy number resemble euploid expectations. LD-PGTA leverages knowledge of allele frequencies and linkage disequilibrium in external phased population reference panels and contrasts the likelihood of the observed data under alternative ploidy hypotheses. We previously applied this method retrospectively to low-coverage WGS-based PGT-A data, uncovering evidence of potential ploidy abnormalities and establishing proof of principle that requires validation with orthogonal evidence and clinical outcomes. Study design, size, duration This retrospective validation study analyzed a training cohort collected between April 2020 and August 2022. The dataset was obtained from CReATe Fertility Centre (Toronto, Canada) and comprised 179 embryo biopsies that underwent WGS at ∼0.03× coverage. The training cohort included 100 euploid, 67 triploid, and 12 haploid cases, assigned to these categories based on orthogonal evidence from short tandem repeat genotyping. Samples suspected of genome-wide ploidy abnormalities but lacking STR confirmation were excluded. Participants/materials, setting, methods Embryos were classified as haploid or triploid when at least eight autosomes were called monosomic or trisomic, respectively, and the 95% confidence interval of the log-likelihood ratio (LLR) did not span zero. For haploid classification, LLRs were aggregated across entire chromosomes. For triploid classification, LLRs favoring BPH (bi-parental-haplotypes) were aggregated, and the total length of these windows was required to exceed a dynamic threshold of 5–30Mb, depending on autosome length, to be called trisomic. Main results and the role of chance For triploidy detection, LD-PGTA achieved a sensitivity of 82% (49/60) at 100% specificity (91/91). For haploidy detection, sensitivity was 75% (9/12) with 99% specificity (93/94). Triploidy prediction accuracy did not differ between embryos sequenced below 0.05× and those sequenced at ≥ 0.05× coverage (Fisher’s exact test, p = 1.00), with comparable false prediction rates (7.7% vs. 5.9%).%). Given that LD-PGTA leverages knowledge of haplotypes from reference panels and that patterns of linkage disequilibrium vary across populations, it is important to consider performance with regard to the genetic ancestry of target samples. While most embryos (72%; 125/174) exhibited genetic similarity to reference individuals from European populations, several embryos also exhibited genetic similarity to reference individuals from other global populations, underscoring the method’s robustness in multi-ancestry cohorts. Triploidy prediction accuracy did not differ significantly between European and non-European embryos (Fisher’s exact test, p = 0.18), underscoring the method’s robustness in multi-ancestry cohorts. Together, these results demonstrate robust discrimination of genome-wide ploidy errors using haplotype-based inference at low sequencing depth. Limitations, reasons for caution Results are derived from a training dataset and confirmation in an independent validation cohort is underway. Rare ploidy states and mosaicism may reduce sensitivity. Wider implications of the findings Although 2PN embryos are predominantly diploid, ploidy abnormalities occur in ∼0.7–1.8% of 2PN blastocysts. Incorporating LD-PGT-A into routine workflows may prevent transfer of haploid or triploid embryos and rescue some 3PN and 0/1PN embryos, increasing the pool of usable embryos and improving diagnostic accuracy of PGT-A without increased sequencing depth. Trial registration number No