Structural variation across 138,134 samples in the TOPMed consortium

Abstract

Ever larger Structural Variant (SV) catalogs highlighting the diversity within and between populations help researchers better understand the links between SVs and disease. The identification of SVs from DNA sequence data is non-trivial and requires a balance between comprehensiveness and precision. Here we present a catalog of 355,667 SVs (59.34% novel) across autosomes and the X chromosome (50bp+) from 138,134 individuals in the diverse TOPMed consortium. We describe our methodologies for SV inference resulting in high variant quality and >90% allele concordance compared to long-read de-novo assemblies of well-characterized control samples. We demonstrate utility through significant associations between SVs and important various cardio-metabolic and hemotologic traits. We have identified 690 SV hotspots and deserts and those that potentially impact the regulation of medically relevant genes. This catalog characterizes SVs across multiple populations and will serve as a valuable tool to understand the impact of SV on disease development and progression.

Keywords: NGS; Population; Structural Variants; TOPMed.

Figure 1:

Overview of SV calls. A) Sample counts based on genetically inferred ancestry showing the majority of individuals are Europeans (EUR) followed by African (AFR), East Asian and Samoan (EAS), American (AMR), South Asian (SAS) and Mestizo (MES) ancestry. B) Per-sample SV count distributions by ancestry. C) Overview of gene density (red), deletions (blue), duplications (orange) and inversions (green). D) Size distribution of population genotyped CNV and inversions. The majority of SVs across the population are large events. E) Randomized PCA principal components 1 and 2 of deletions. See supplementary Figures 2-4 for deletions , duplications and inversions.

Figure 2:

Evaluation of SV call sets against using haplotype-resolved assemblies for deletion (DEL) and duplication (DUP) calls in the NA12878 and NA19238 genomes. A) Evaluation using TT-Mars. The fraction of calls that may be assessed using the assemblies (analyzed) and positive predictive value (PPV) are given for DEL and DUP calls. B) Support for calls from both the TT-Mars and Truvari methods. C) The size by count spectrum of all calls (red), the count validated by TT-Mars (green), and the count validated by Truvari (blue) for the combination of both genomes.

Figure 3:

A) Assessment of Mendelian error and novel HET rates per SV type across 11,387 samples from trio/duo families. For the evaluation, we also include the multiallelic variants that do increase the error rate, especially across duplications. B) Overlap of SV over 1KGP and gnomAD SV with respect to the allele frequency within the TOPMed SV call set. The allele frequencies change slightly between the novel SV from TOPMed and other overlapping SV. C) FST plot of African versus European ancestry of SV across the entire genome, highlighting a threshold of 0.11.

Figure 4:. Overview of the impact of the SVs and their clustering along the genome.

A) SVs identified and clustered based on their haploinsufficiency (HI) and triplosensitivity (TriS) potential across different allele counts. B) Overview of SV hotspots and deserts across the TOPMed cohort. Here deserts are regions of the genome with no SV identifiable despite the large collection of individuals in this study.