Phasing for medical sequencing using rare variants and large haplotype reference panels

Inferring compound heterozygosity from large-scale exome sequencing data

Recessive diseases arise when both copies of a gene are impacted by a damaging genetic variant. When a patient carries two potentially causal variants in a gene, accurate diagnosis requires determining that these variants occur on different copies of the chromosome (that is, are in trans) rather than on the same copy (that is, in cis). However, current approaches for determining phase, beyond parental testing, are limited in clinical settings. Here we developed a strategy for inferring phase for rare variant pairs within genes, leveraging genotypes observed in the Genome Aggregation Database (v2, n = 125,748 exomes). Our approach estimates phase with 96% accuracy, both in trio data and in patients with Mendelian conditions and presumed causal compound heterozygous variants. We provide a public resource of phasing estimates for coding variants and counts per gene of rare variants in trans that can aid interpretation of rare co-occurring variants in the context of recessive disease.

cLD: Rare-variant linkage disequilibrium between genomic regions identifies novel genomic interactions

Linkage disequilibrium (LD) is a fundamental concept in genetics; critical for studying genetic associations and molecular evolution. However, LD measurements are only reliable for common genetic variants, leaving low-frequency variants unanalyzed. In this work, we introduce cumulative LD (cLD), a stable statistic that captures the rare-variant LD between genetic regions, which reflects more biological interactions between variants, in addition to lack of recombination. We derived the theoretical variance of cLD using delta methods to demonstrate its higher stability than LD for rare variants. This property is also verified by bootstrapped simulations using real data. In application, we find cLD reveals an increased genetic association between genes in 3D chromatin interactions, a phenomenon recently reported negatively by calculating standard LD between common variants. Additionally, we show that cLD is higher between gene pairs reported in interaction databases, identifies unreported protein-protein interactions, and reveals interacting genes distinguishing case/control samples in association studies.

Inferring compound heterozygosity from large-scale exome sequencing data

Recessive diseases arise when both the maternal and the paternal copies of a gene are impacted by a damaging genetic variant in the affected individual. When a patient carries two different potentially causal variants in a gene for a given disorder, accurate diagnosis requires determining that these two variants occur on different copies of the chromosome (i.e., are in trans) rather than on the same copy (i.e. in cis). However, current approaches for determining phase, beyond parental testing, are limited in clinical settings. We developed a strategy for inferring phase for rare variant pairs within genes, leveraging genotypes observed in exome sequencing data from the Genome Aggregation Database (gnomAD v2, n=125,748). When applied to trio data where phase can be determined by transmission, our approach estimates phase with 95.7% accuracy and remains accurate even for very rare variants (allele frequency < 1×10-4). We also correctly phase 95.9% of variant pairs in a set of 293 patients with Mendelian conditions carrying presumed causal compound heterozygous variants. We provide a public resource of phasing estimates from gnomAD, including phasing estimates for coding variants across the genome and counts per gene of rare variants in trans, that can aid interpretation of rare co-occurring variants in the context of recessive disease.

Genotyping, the Usefulness of Imputation to Increase SNP Density, and Imputation Methods and Tools

Imputation has become a standard practice in modern genetic research to increase genome coverage and improve accuracy of genomic selection and genome-wide association study as a large number of samples can be genotyped at lower density (and lower cost) and, imputed up to denser marker panels or to sequence level, using information from a limited reference population. Most genotype imputation algorithms use information from relatives and population linkage disequilibrium. A number of software for imputation have been developed originally for human genetics and, more recently, for animal and plant genetics considering pedigree information and very sparse SNP arrays or genotyping-by-sequencing data. In comparison to human populations, the population structures in farmed species and their limited effective sizes allow to accurately impute high-density genotypes or sequences from very low-density SNP panels and a limited set of reference individuals. Whatever the imputation method, the imputation accuracy, measured by the correct imputation rate or the correlation between true and imputed genotypes, increased with the increasing relatedness of the individual to be imputed with its denser genotyped ancestors and as its own genotype density increased. Increasing the imputation accuracy pushes up the genomic selection accuracy whatever the genomic evaluation method. Given the marker densities, the most important factors affecting imputation accuracy are clearly the size of the reference population and the relationship between individuals in the reference and target populations.

Haplotype-resolved germline and somatic alterations in renal medullary carcinomas

BACKGROUND

Renal medullary carcinomas (RMCs) are rare kidney cancers that occur in adolescents and young adults of African ancestry. Although RMC is associated with the sickle cell trait and somatic loss of the tumor suppressor, SMARCB1, the ancestral origins of RMC remain unknown. Further, characterization of structural variants (SVs) involving SMARCB1 in RMC remains limited.

METHODS

We used linked-read genome sequencing to reconstruct germline and somatic haplotypes in 15 unrelated patients with RMC registered on the Children's Oncology Group (COG) AREN03B2 study between 2006 and 2017 or from our prior study. We performed fine-mapping of the HBB locus and assessed the germline for cancer predisposition genes. Subsequently, we assessed the tumor samples for mutations outside of SMARCB1 and integrated RNA sequencing to interrogate the structural variants at the SMARCB1 locus.

RESULTS

We find that the haplotype of the sickle cell mutation in patients with RMC originated from three geographical regions in Africa. In addition, fine-mapping of the HBB locus identified the sickle cell mutation as the sole candidate variant. We further identify that the SMARCB1 structural variants are characterized by blunt or 1-bp homology events.

CONCLUSIONS

Our findings suggest that RMC does not arise from a single founder population and that the HbS allele is a strong candidate germline allele which confers risk for RMC. Furthermore, we find that the SVs that disrupt SMARCB1 function are likely repaired by non-homologous end-joining. These findings highlight how haplotype-based analyses using linked-read genome sequencing can be applied to identify potential risk variants in small and rare disease cohorts and provide nucleotide resolution to structural variants.

Impact of pre- and post-variant filtration strategies on imputation

Quality control (QC) methods for genome-wide association studies and fine mapping are commonly used for imputation, however they result in loss of many single nucleotide polymorphisms (SNPs). To investigate the consequences of filtration on imputation, we studied the direct effects on the number of markers, their allele frequencies, imputation quality scores and post-filtration events. We pre-phrased 1031 genotyped individuals from diverse ethnicities and compared the imputed variants to 1089 NCBI recorded individuals for additional validation. Without QC-based variant pre-filtration, we observed no impairment in the imputation of SNPs that failed QC whereas with pre-filtration there was an overall loss of information. Significant differences between frequencies with and without pre-filtration were found only in the range of very rare (5E-04-1E-03) and rare variants (1E-03-5E-03) (p < 1E-04). Increasing the post-filtration imputation quality score from 0.3 to 0.8 reduced the number of single nucleotide variants (SNVs) < 0.001 2.5 fold with or without QC pre-filtration and halved the number of very rare variants (5E-04). Thus, to maintain confidence and enough SNVs, we propose here a two-step filtering procedure which allows less stringent filtering prior to imputation and post-imputation in order to increase the number of very rare and rare variants compared to conservative filtration methods.

Accurate, scalable and integrative haplotype estimation

The number of human genomes being genotyped or sequenced increases exponentially and efficient haplotype estimation methods able to handle this amount of data are now required. Here we present a method, SHAPEIT4, which substantially improves upon other methods to process large genotype and high coverage sequencing datasets. It notably exhibits sub-linear running times with sample size, provides highly accurate haplotypes and allows integrating external phasing information such as large reference panels of haplotypes, collections of pre-phased variants and long sequencing reads. We provide SHAPEIT4 in an open source format and demonstrate its performance in terms of accuracy and running times on two gold standard datasets: the UK Biobank data and the Genome In A Bottle.

BCFtools/csq: haplotype-aware variant consequences

Motivation

Prediction of functional variant consequences is an important part of sequencing pipelines, allowing the categorization and prioritization of genetic variants for follow up analysis. However, current predictors analyze variants as isolated events, which can lead to incorrect predictions when adjacent variants alter the same codon, or when a frame-shifting indel is followed by a frame-restoring indel. Exploiting known haplotype information when making consequence predictions can resolve these issues.

Results

BCFtools/csq is a fast program for haplotype-aware consequence calling which can take into account known phase. Consequence predictions are changed for 501 of 5019 compound variants found in the 81.7M variants in the 1000 Genomes Project data, with an average of 139 compound variants per haplotype. Predictions match existing tools when run in localized mode, but the program is an order of magnitude faster and requires an order of magnitude less memory.

Availability and Implementation

The program is freely available for commercial and non-commercial use in the BCFtools package which is available for download from http://samtools.github.io/bcftools.

Supplementary information

Supplementary data are available at Bioinformatics online.

Genetic analysis of deep phenotyping projects in common disorders

Several studies of complex psychotic disorders with large numbers of neurobiological phenotypes are currently under way, in living patients and controls, and on assemblies of brain specimens. Genetic analyses of such data typically present challenges, because of the choice of underlying hypotheses on genetic architecture of the studied disorders and phenotypes, large numbers of phenotypes, the appropriate multiple testing corrections, limited numbers of subjects, imputations required on missing phenotypes and genotypes, and the cross-disciplinary nature of the phenotype measures. Advances in genotype and phenotype imputation, and in genome-wide association (GWAS) methods, are useful in dealing with these challenges. As compared with the more traditional single-trait analyses, deep phenotyping with simultaneous genome-wide analyses serves as a discovery tool for previously unsuspected relationships of phenotypic traits with each other, and with specific molecular involvements.

Strategies for phasing and imputation in a population isolate

In the search for genetic associations with complex traits, population isolates offer the advantage of reduced genetic and environmental heterogeneity. In addition, cost-efficient next-generation association approaches have been proposed in these populations where only a subsample of representative individuals is sequenced and then genotypes are imputed into the rest of the population. Gene mapping in such populations thus requires high-quality genetic imputation and preliminary phasing. To identify an effective study design, we compare by simulation a range of phasing and imputation software and strategies. We simulated 1,115,604 variants on chromosome 10 for 477 members of the large complex pedigree of Campora, a village within the established isolate of Cilento in southern Italy. We assessed the phasing performance of identical by descent based software ALPHAPHASE and SLRP, LD-based software SHAPEIT2, SHAPEIT3, and BEAGLE, and new software EAGLE that combines both methodologies. For imputation we compared IMPUTE2, IMPUTE4, MINIMAC3, BEAGLE, and new software PBWT. Genotyping errors and missing genotypes were simulated to observe their effects on the performance of each software. Highly accurate phased data were achieved by all software with SHAPEIT2, SHAPEIT3, and EAGLE2 providing the most accurate results. MINIMAC3, IMPUTE4, and IMPUTE2 all performed strongly as imputation software and our study highlights the considerable gain in imputation accuracy provided by a genome sequenced reference panel specific to the population isolate.

Reference-based phasing using the Haplotype Reference Consortium panel

Haplotype phasing is a fundamental problem in medical and population genetics. Phasing is generally performed via statistical phasing in a genotyped cohort, an approach that can yield high accuracy in very large cohorts but attains lower accuracy in smaller cohorts. Here we instead explore the paradigm of reference-based phasing. We introduce a new phasing algorithm, Eagle2, that attains high accuracy across a broad range of cohort sizes by efficiently leveraging information from large external reference panels (such as the Haplotype Reference Consortium; HRC) using a new data structure based on the positional Burrows-Wheeler transform. We demonstrate that Eagle2 attains a ∼20× speedup and ∼10% increase in accuracy compared to reference-based phasing using SHAPEIT2. On European-ancestry samples, Eagle2 with the HRC panel achieves >2× the accuracy of 1000 Genomes-based phasing. Eagle2 is open source and freely available for HRC-based phasing via the Sanger Imputation Service and the Michigan Imputation Server.

A reference panel of 64,976 haplotypes for genotype imputation

We describe a reference panel of 64,976 human haplotypes at 39,235,157 SNPs constructed using whole-genome sequence data from 20 studies of predominantly European ancestry. Using this resource leads to accurate genotype imputation at minor allele frequencies as low as 0.1% and a large increase in the number of SNPs tested in association studies, and it can help to discover and refine causal loci. We describe remote server resources that allow researchers to carry out imputation and phasing consistently and efficiently.

Haplotype estimation for biobank-scale data sets

The UK Biobank (UKB) has recently released genotypes on 152,328 individuals together with extensive phenotypic and lifestyle information. We present a new phasing method SHAPEIT3 that can handle such biobank scale datasets and results in switch error rates as low as ~0.3%. The method exhibits O(NlogN) scaling in sample size (N), enabling fast and accurate phasing of even larger cohorts.