A novel method for detecting uniparental disomy from trio genotypes identifies a significant excess in children with developmental disorders

Abstract

Exome sequencing of parent-offspring trios is a popular strategy for identifying causative genetic variants in children with rare diseases. This method owes its strength to the leveraging of inheritance information, which facilitates de novo variant calling, inference of compound heterozygosity, and the identification of inheritance anomalies. Uniparental disomy describes the inheritance of a homologous chromosome pair from only one parent. This aberration is important to detect in genetic disease studies because it can result in imprinting disorders and recessive diseases. We have developed a software tool to detect uniparental disomy from child-mother-father genotype data that uses a binomial test to identify chromosomes with a significant burden of uniparentally inherited genotypes. This tool is the first to read VCF-formatted genotypes, to perform integrated copy number filtering, and to use a statistical test inherently robust for use in platforms of varying genotyping density and noise characteristics. Simulations demonstrated superior accuracy compared with previously developed approaches. We implemented the method on 1057 trios from the Deciphering Developmental Disorders project, a trio-based rare disease study, and detected six validated events, a significant enrichment compared with the population prevalence of UPD (1 in 3500), suggesting that most of these events are pathogenic. One of these events represents a known imprinting disorder, and exome analyses have identified rare homozygous candidate variants, mainly in the isodisomic regions of UPD chromosomes, which, among other variants, provide targets for further genetic and functional evaluation.

Figure 1.

Study workflow. The study consisted of three main steps: data preparation, UPD detection, and candidate variant analysis. In the data preparation stage, we collected informative genotypes seen in all members of each trio. Either a multi-sample trio VCF or three single-sample VCFs can be used as input; the latter requires the annotation of homozygous reference genotypes, not usually encoded in single-sample VCF files. In the UPD detection stage, we selected trios containing a proband chromosome with an enrichment of UPD-informative genotypes. Exomes (five) available for samples with a detected UPD event were selected for the candidate workup analysis, in which we attempted to find rare protein-altering variants that may manifest in the proband’s phenotypes.

Figure 2.

Sensitivity of UPD detection simulations. Simulations to assess sensitivity of UPD detections at different sizes, from different data sources. (iUPD) Isodisomy; (hUPD) heterodisomy.

Figure 3.

Sensitivity comparisons. Simulations were performed to measure the sensitivity of detecting introduced UPD events from SNP and exome data, ranging in size from 1 Mb to chromosomal.

Figure 4.

Specificity comparisons. Simulations on normal SNP and exome samples were compared to measure the proportion of samples without UPD detections.

Figure 5.

Receiver operator characteristic curve comparing UPD detection accuracy. Accuracy of UPD detection at different simulated UPD sizes. (dio) UPDio; (tool) UPDtool; (trio) SNPtrio.

Figure 6.

DDD UPD P-value distributions. Distribution of the –log₁₀ P-values for UPD detections from different data sources, with or without CNV data. Presence of sample-specific CNV data increases the proportion of extremely significant events and decreases the proportion of marginally significant events. P-value minimum truncated to 1 × 10⁻¹⁰⁰.

Figure 7.

UPD example plot. A plot of QC-passing proband genotypes on each autosome. The position and color reflect zygosity (homozygous, heterozygous) and informative state (biparental inheritance, maternal isodisomy, maternal heterodisomy or isodisomy, paternal isodisomy, paternal heterodisomy or isodisomy). The figure displays each chromosome ideogram. Each chromosome has an x-axis (chromosome position) and y-axis (zygosity, and informative event type). In this case, the UPD event for chromosome 2 is depicted with a mixture of dark-green points (maternal isodisomy) and light-green points (maternal isodisomy or maternal heterodisomy). The zygosity row demonstrates homozygosity along the entirety of the chromosome, reflecting the complete isodisomy.