Haplotype-aware inference of human chromosome abnormalities

Abstract

Extra or missing chromosomes-a phenomenon termed aneuploidy-frequently arise during human meiosis and embryonic mitosis and are the leading cause of pregnancy loss, including in the context of in vitro fertilization (IVF). While meiotic aneuploidies affect all cells and are deleterious, mitotic errors generate mosaicism, which may be compatible with healthy live birth. Large-scale abnormalities such as triploidy and haploidy also contribute to adverse pregnancy outcomes, but remain hidden from standard sequencing-based approaches to preimplantation genetic testing for aneuploidy (PGT-A). The ability to reliably distinguish meiotic and mitotic aneuploidies, as well as abnormalities in genome-wide ploidy, may thus prove valuable for enhancing IVF outcomes. Here, we describe a statistical method for distinguishing these forms of aneuploidy based on analysis of low-coverage whole-genome sequencing data, which is the current standard in the field. Our approach overcomes the sparse nature of the data by leveraging allele frequencies and linkage disequilibrium (LD) measured in a population reference panel. The method, which we term LD-informed PGT-A (LD-PGTA), retains high accuracy down to coverage as low as 0.05 × and at higher coverage can also distinguish between meiosis I and meiosis II errors based on signatures spanning the centromeres. LD-PGTA provides fundamental insight into the origins of human chromosome abnormalities, as well as a practical tool with the potential to improve genetic testing during IVF.

Keywords: haploidy; in vitro fertilization; mosaicism; triploidy; trisomy.

Fig. 1.

Signatures of various forms of chromosome abnormality with respect to their composition of identical and distinct parental homologs. Normal gametogenesis produces two genetically distinct copies of each chromosome—one copy from each parent—that comprise mosaics of two homologs possessed by each parent. Meiotic-origin trisomies may be diagnosed by the presence of one or more tracts with three distinct parental homologs (i.e., transmission of BPH from a given parent). In contrast, mitotic-origin trisomies are expected to exhibit only two genetically distinct parental homologs chromosome-wide (i.e., duplication of a SPH from a given parent). Triploidy and haploidy will mirror patterns observed for individual meiotic trisomies and monosomies, respectively, but across all 23 chromosome pairs—a pattern that confounds standard coverage-based analysis of PGT-A data.

Fig. 2.

(A) Within defined genomic windows, select reads overlapping informative SNPs that tag common haplotype variation. (B) Obtain joint frequencies of corresponding SNPs from a phased panel of ancestry-matched reference haplotypes. (C) Randomly resample 2 to 18 reads and compute probabilities of observed alleles under two competing ploidy hypotheses. (D) Compare the hypotheses by computing a likelihood ratio and estimate the mean and variance by resampling random sets of reads using a bootstrapping approach. (E) Repeat steps A–D for consecutive nonoverlapping genomic windows and aggregate the log-likelihood ratios over larger genomic intervals. (F) Use the mean and variance of the combined log-likelihood ratio to estimate a confidence interval and classify the ploidy state of the genomic interval.

Fig. 3.

Balanced ROC curves for BPH vs. SPH with matched and random reference panels for nonadmixed embryos, varying depths of coverage. Each balanced ROC curve reflects an average over bins across the genome. We averaged both the BTPR and BFPR for common z scores across bins.

Fig. 4.

Demonstration of the detection of meiotic cross-overs from low-coverage PGT-A data. Trisomies were simulated with varying locations of meiotic cross-overs, as depicted in Top diagrams, and varying depths of coverage (0.01×, 0.05×, and 0.1×). Confidence intervals correspond to a z score of one (confidence level of 68.3%). The size of the genomic windows (GW) varies with the coverage, while the size of the bins is kept constant.

Fig. 5.

Ancestry inference from low-coverage PGT-A data informs the selection of matched reference panels. Principal component axes were defined based on analysis of 1000 Genomes reference samples (A and C) and colored according to superpopulation annotations (African [AFR], Admixed American [AMR], East Asian [EAS], European [EUR], South Asian [SAS]). Low-coverage embryo samples were then projected onto these axes using a Procrustes approach implemented with LASER (v2.0) (50) (B and D) and their ancestries were classified using k-nearest neighbors of the first four principal component axes (k = 10). Projections of first-generation admixed reference samples were approximated as the mean of random samples from each of the component superpopulations. A and B depict principal components 1 and 2, while C and D depict principal components 2 and 3, together capturing continental-scale patterns of global ancestry.

Fig. 6.

Application of LD-PGTA to low-coverage sequencing-based data. (A) LD-PGTA was used to refine classification results from 2,988 chromosomes originally diagnosed as whole-chromosome (i.e., nonsegmental) aneuploid based on standard coverage-based analysis with Bluefuse Multi, including strong validation of putative monosomies, as well as subclassification of BPH and SPH trisomies. (B) Association of BPH and SPH trisomies with maternal age. (C) LD-PGTA classifications of BPH versus SPH trisomy, stratifying by individual chromosome. BPH trisomy is strongly enriched on chromosomes 16 and 22, while signatures of SPH trisomy are more evenly distributed among the various autosomes. Chromosomes were classified using a 50% confidence interval. Dashed line indicates x = y.

Fig. 7.

Visualization of results from representative putative diploid (A), triploid (B), and haploid (C) samples. Copy number estimates obtained using a standard coverage-based approach (BlueFuse Multi) are depicted in A–C, Left and are indicative of diploidy. LD-PGTA results are depicted in A–C, Center and Right and suggest genome-wide abnormalities in ploidy for the latter two samples.