01 Briefings in Bioinformatics · 1 September 2026_
Comprehensive evaluation of structural variation detection for germline and somatic analysis with long-read sequencing data
- Structural variation
- Methods
- Cancer genomics
14 long-read SV callers on 20 PacBio CLR, HiFi and ONT datasets: DeBreak, cuteSV2 and SVDF lead for germline calls, Severus and SAVANA for somatic, nanomonsv at low coverage.
Summary
The authors built one benchmarking framework for germline and somatic structural variant detection and ran 14 callers across 20 core datasets from three long-read platforms. Performance was scored on 12 dimensions, including baseline artefact rate and two Mendelian consistency measures from trios. For germline SVs, DeBreak, cuteSV2 and SVDF were accurate and stable; cuteSV2 and SVHunter kept genotyping accuracy across depths; Severus and cuteSV2 handled inversions, duplications and translocations best. In tumours, somatic-specific callers beat germline callers. Code and resources are public.
What to take from it
- Germline: DeBreak, cuteSV2, SVDF; genotyping across depth: cuteSV2, SVHunter
- Complex SVs (INV, DUP, TRA): Severus and cuteSV2
- Somatic: use a somatic caller - Severus, SAVANA; nanomonsv at low coverage
- Mendelian error rates used as a truth-free accuracy check
Review
Useful as a current caller shortlist, and the Mendelian discordance metrics give an accuracy check that does not depend on a truth set - the situation for any livestock trio. The limits: rankings depend on the truth sets and matching parameters used for evaluation, and SV benchmarks are known to shift with breakpoint tolerance and size thresholds. All data are human, on a linear reference; repeat-rich and segmentally duplicated regions of other species will stress callers differently, and no graph-based genotyping is compared. Before adopting a caller from this, check the per-SV-type and per-platform tables for your own read type and depth, and rerun the public pipeline on one of your own trios.
02 Nature Communications · 29 August 2026_
Deviations from genetic additivity driven by rare variants at biobank scale
- Human genetics
- Methods
An orthogonal allele recoding tests dominance without Hardy-Weinberg assumptions inside standard mixed models; across 2906 plasma proteins in up to 399,943 UK Biobank people, about a third of cis gene-protein links show non-linear dose response.
Summary
Existing tests for non-additive effects either do not scale or assume Hardy-Weinberg equilibrium, which fails for rare variants. The authors recode genotypes into orthogonal additive and dominance components that plug into existing linear mixed models. Applied to 2906 plasma proteins and 55 quantitative traits, they find non-linear dose response in about one third of cognate gene-protein pairs, recessive effects such as FUT10 variants on lung function (replicated in All of Us), and partial recessivity across many complex traits. A joint two-degree-of-freedom test gains power when a trait is partially recessive.
What to take from it
- Dominance test that does not assume Hardy-Weinberg, so it works for rare variants
- ~1/3 of cis gene-protein pairs non-linear in dose
- Recessive FUT10 effect on lung function, replicated in All of Us
- 2-df additive plus dominance test gains power under partial recessivity
Review
The method fixes a practical problem: dominance terms for rare variants are unstable when coded under Hardy-Weinberg assumptions. It matters for livestock, where inbreeding and structured mating make departures from Hardy-Weinberg normal and recessive effects on fertility and viability are the main interest. The weak points: dominance signals for rare variants rest on very few homozygotes, so a handful of individuals can drive a result, and any residual population structure or relatedness not captured by the model can mimic non-additivity. Plasma protein levels also have assay artefacts (epitope effects of coding variants) that can create apparent dose non-linearity. Before using the recessive calls, ask for homozygote counts per hit, sensitivity to excluding related individuals, and replication beyond the single FUT10 example.
03 Cell Genomics · 29 September 2026_
Disease-associated loci share properties with response eQTLs under common environmental exposures
- Human genetics
Differentiating cultures from 51 iPSC lines exposed to nicotine, caffeine or ethanol (1.4 million cells) yield hundreds of response eQTLs that look like disease loci - distal enhancers, constrained genes - and overlap GWAS signals that baseline eQTLs miss.
Summary
Many GWAS loci are not explained by eQTLs from healthy adult tissues. The authors exposed heterogeneous differentiating cultures from 51 human iPSC lines to nicotine, caffeine or ethanol and profiled 1.4 million cells by single-cell RNA-seq. They found hundreds of response eQTLs across cell types and effects that change along differentiation. Unlike steady-state eQTLs, and like disease loci, response eQTLs sit in distal enhancers, regulate genes under strong constraint and have complex regulatory landscapes. They also overlap disease loci not explained by baseline eQTLs.
What to take from it
- 51 iPSC lines x 3 exposures, 1.4M cells, hundreds of response eQTLs
- Response eQTLs resemble GWAS loci; baseline eQTLs do not
- Some GWAS loci explained only under exposure
- Supports mapping eQTLs under stimulation, not only at steady state
Review
The result gives a concrete reason for the gap between GWAS hits and eQTL colocalisation, and the design - many cell types from one differentiating culture, several exposures - is efficient. The limits are sample size and relevance. Fifty-one donors gives low power per cell type, so the response eQTL set is enriched for large effects, and the comparison with steady-state eQTLs may partly reflect differences in power rather than biology. Exposure doses and in vitro cell states may not match what happens in tissues. The overlap with disease loci needs formal colocalisation with fine-mapping, not proximity. For livestock, the lesson is that eQTL atlases from unstimulated tissues will miss regulatory variants that act under infection, heat or diet challenges.
04 Journal of Animal Science · 29 September 2026_
GWAS on imputed whole-genome sequence variants reveal novel candidate mutations affecting feed efficiency and body composition traits in a crossbred swine population
- Animal science
Single-step GWAS on sequence imputed from 50-60K arrays in 7,339 genotyped crossbred pigs finds 133 associated LD blocks for feed efficiency and body composition, with predicted-damaging variants in SIRT2, RYR1, MC4R and LEPR.
Summary
The authors ran single-step GWAS with whole-genome sequence variants imputed from 50-60K SNP arrays in a three-generation crossbred swine pedigree of 14,242 pigs (9,336 phenotyped, 7,339 genotyped). Variants were grouped into 705,923 LD blocks and tested for feed efficiency and body composition traits. 133 blocks were significant, some with pleiotropic effects, spanning 453 candidate genes. Four genes carried moderate or high impact variants: SIRT2, RYR1, MC4R and LEPR.
What to take from it
- 14,242-pig pedigree; sequence imputed from 50-60K arrays
- 133 significant LD blocks, 453 candidate genes
- Coding candidates: SIRT2, RYR1, MC4R, LEPR
- Testing LD blocks rather than single SNPs
Review
The data set is large for swine feed efficiency, and recovering MC4R, LEPR and RYR1 - genes with known effects on growth, fat and meat quality in pigs - shows the pipeline works. What it does not show is new causal mutations. Imputing sequence from a 50-60K array loses accuracy for low-frequency variants, which are exactly the ones that would carry new causal alleles, and imputation accuracy per variant is not given in the abstract. Moderate-impact coding variants in large LD blocks are candidates only; the MC4R and RYR1 variants may simply be the known ones. Before using any of the 133 blocks for selection, ask for imputation accuracy against sequenced animals, conditional analyses on the known MC4R and RYR1 alleles, and validation in a separate population.
05 Animal Genetics · 1 October 2026_
Weighted single-step GWAS reveals genetic mechanisms underlying pregnancy loss in early Nellore heifers
- Animal science
In 23,507 Nellore heifers (17,052 genotyped), pregnancy loss from diagnosis to calving has heritability 0.033; 25 windows explain 31% of additive variance, the largest on BTA16 with no annotated gene.
Summary
Pregnancy loss was defined as failure between pregnancy diagnosis and calving in early-bred Nellore heifers. The authors ran weighted single-step GWAS on 23,507 heifers, 17,052 genotyped on about 44,000 SNPs. Heritability was 0.033. Twenty-five genomic regions on 16 chromosomes explained 31.16% of additive genetic variance; the top region on BTA16 explained 6.51% but contains no annotated protein-coding gene. Candidates include IL4R, IL21R, LRRC4C, RBMS3, ZFPM2, PRSS23 and NDST4, with enrichment for immune, adhesion, chromatin and WNT pathways.
What to take from it
- h2 = 0.033 for pregnancy loss in Nellore heifers
- 25 windows explain 31% of additive variance - window-based, not per-SNP significance
- Top region BTA16 (6.5%) has no annotated protein-coding gene
- Candidate genes are immune and adhesion related
Review
The phenotype is economically important and rarely studied in Bos indicus, and the sample is large. The weak point is statistical: with heritability of 0.033 on a binary trait, the share of additive variance attributed to each window in weighted ssGWAS is noisy, and the method reports windows above a variance threshold rather than tests with controlled error rates. Twenty-five windows explaining 31% of a variance component that small should be read as a ranked list, not a set of confirmed loci. The candidate genes are chosen by proximity and pathway plausibility. Before acting on any region, ask for a replication sample or cross-validation of window rankings, and for a test of whether the BTA16 region holds a non-coding or structural feature that the SNP panel tags.