01 Genome Research · 16 September 2026_
SwinePan for pig graph-based pangenome and multiomics data mining
- Animal science
- Structural variation
- Methods
A graph-based pig pangenome covering 2,598 pigs across 35 breeds, packaged with GWAS and eQTL tools and over 150,000 phenotypes so structural variants can be tied straight to traits.
Summary
Single linear pig reference genomes miss structural variants (SVs) that differ between breeds. SwinePan builds a graph-based pangenome from genomes spanning pig diversity, then genotypes 2,598 individuals across 35 breeds against it: 185,759 SVs, 117 million SNPs, 6.8 million indels. It layers on transcriptomes from liver, loin muscle, abdominal fat and backfat, plus over 150,000 phenotypic records, and ships an online toolkit for GWAS, eQTL mapping and colocalization. Two worked examples show it locating trait-associated loci and pointing at plausible regulatory mechanisms behind them.
What to take from it
- 2,598 pigs, 35 breeds, graph-based pangenome: 185,759 SVs, 117M SNPs, 6.8M indels
- Multiomics layer: four tissue transcriptomes plus 150,000+ phenotype records
- Built-in GWAS, eQTL mapping and colocalization toolkit, not just a variant catalogue
- Two proof-of-concept analyses tie SVs to trait loci and candidate regulatory mechanisms
Review
The value here is the database and toolkit, not a single finding: it turns a graph pangenome into something a breeder or geneticist can query without building their own pipeline. The two worked examples are described as proof-of-concept, not validated causal mechanisms, so the highlighted loci are leads, not confirmed causal variants until followed up with independent functional evidence. It is unclear from the abstract how the 35 breeds are weighted or whether rarer breeds have enough depth for reliable SV calling, which matters for anyone wanting to use this outside the sampled breeds. Worth checking before relying on it: SV calling accuracy at breed-specific loci, and whether the toolkit corrects properly for population structure across such a broad breed panel.
02 Nature Genetics · 15 September 2026_
Pan-genome-based resequencing of 2,320 accessions reveals structural variations and accelerates breeding advances in cultivated peanut
- Structural variation
- Population genetics
- Methods
A graph-based pan-genome from 14 peanut genomes, genotyped across 2,320 accessions, resolves the genome-doubling exchanges that block peanut breeding and is used to build new high-yield dwarf lines.
Summary
Cultivated peanut is allotetraploid, and homoeologous exchanges between its two subgenomes confuse genotyping against a single reference, slowing breeding despite peanut's importance as a global legume crop. The authors built a graph-based pan-genome from 14 genomes spanning all six peanut varieties, then genotyped 2,320 accessions covering most of the ICRISAT and USDA core germplasm collections. The pan-genome resolves structural variation and homoeologous exchange calls that a linear reference gets wrong, and the team used it to find genes tied to flowering time and dwarfism. Combining superior haplotypes identified this way with germplasm selection, they report new high-yield dwarf breeding lines.
What to take from it
- Graph pan-genome from 14 genomes spanning all 6 peanut varieties
- 2,320 accessions genotyped: 88% of ICRISAT and 59% of USDA core germplasm
- Resolves homoeologous exchanges between subgenomes that a linear reference miscalls
- Used to identify flowering and dwarfism genes and to develop new high-yield dwarf breeding lines
Review
This is a strong demonstration of pan-genome utility in an allopolyploid crop, where a single reference is known to be a poor fit. The breeding outcome, new dwarf lines, is an end-to-end result rather than just a catalogue, which is the harder bar to clear. What is not clear from the abstract is how homoeologous exchange calls were validated against an independent method, and how much of the germplasm coverage comes from lines closely related to the 14 reference genomes versus more distant material, since pan-genome resolution tends to degrade for samples far from the reference set. Before extending this approach to other allopolyploid crops, check how well the flowering and dwarfism candidate genes replicate in an independent population, since candidate-gene calls in structured germplasm panels are prone to confounding by relatedness.
03 Genetics Selection Evolution · 7 September 2026_
Impact of genomic selection for disease resistance on the spread of infection in a simulated aquaculture population
- Animal science
- Population genetics
- Methods
Simulating genomic selection for disease resistance in farmed salmon shows that breeding for survival time, not just survival status, is what actually slows disease spread.
Summary
Breeding programs select against mortality using challenge tests, but it has been unclear whether lower mortality also means less transmission, since susceptibility, infectivity and disease-induced mortality are separate underlying traits. The authors simulated 10 generations of genomic selection in a salmon population exposed to a fast-spreading pathogen, modelling infection with a stochastic Susceptible-Infected-Removed framework and varying the genetic correlations between the three underlying traits. Selecting on simple dead/alive status gave transmission outcomes that depended heavily on those correlations and on challenge test design. Selecting on time to death consistently cut the pathogen's basic reproduction number below 1 within four generations, regardless of trait correlations or test design, because it mainly selects for reduced susceptibility rather than just endurance.
What to take from it
- Compares genomic selection on binary mortality versus time-to-death in a simulated salmon population
- Uses a stochastic SIR epidemiological model to separate susceptibility, infectivity and mortality
- Selection on dead/alive status: transmission outcome depends on trait correlations and test design
- Selection on time to death: consistently reduces the reproduction number below 1 within 4 generations
- Direct selection on the underlying epidemiological traits remains the most efficient option, when measurable
Review
The finding is a clean, mechanistic argument for changing what challenge-test data breeding programs record, and the SIR framework is a sensible way to separate the three traits that a plain survival phenotype conflates. Being a simulation, the result is only as good as its assumed genetic architecture and correlation structure; real disease systems may have nonlinearities or genotype-by-environment effects the model does not capture, and the paper does not test robustness against these directly. It is also a single-pathogen, single-species simulation, so the specific numeric result should not be treated as generalizable beyond the modelled scenario. Before applying this to a real breeding program, check whether time-to-death is practically recordable at the scale and cost of existing challenge tests, since the paper's own logic is that measurability is what makes a trait actionable.
04 Frontiers in Veterinary Science · 28 August 2026_
Genome-wide association study reveals candidate genes associated with body weight and wool traits in Ordos fine-wool sheep
- Animal science
- Population genetics
A GWAS in 388 Chinese fine-wool sheep flags six candidate genes, including IGFBP2, IGFBP5 and AXIN1, for body weight and wool traits.
Summary
The Ordos fine-wool sheep is valued in Inner Mongolia for wool quality and meat, but the genetic basis of its body weight and wool traits was not mapped. The authors genotyped 388 animals on a 40K SNP panel and ran a GWAS for body weight, wool length and wool fineness. Twenty-two SNPs reached the suggestive threshold, corresponding to 27 annotated genes; functional annotation narrowed this to six candidates - LAMA2, ARHGAP18, IGFBP2, IGFBP5, CA10 and AXIN1 - as plausible regulators of body weight and wool growth.
What to take from it
- GWAS in 388 Ordos fine-wool sheep using a 40K SNP panel
- 22 suggestive SNPs map to 27 genes for body weight, wool length and wool fineness
- Six candidates highlighted by function: LAMA2, ARHGAP18, IGFBP2, IGFBP5, CA10, AXIN1
- Intended as a starting point for molecular breeding markers in this breed
Review
This is a modest, single-breed GWAS with a sample size, 388, that is small by GWAS standards, and the paper reports suggestive rather than genome-wide-significant associations, so these are candidate loci rather than established ones. IGFBP2, IGFBP5 and AXIN1 have plausible prior biology in growth and Wnt signalling, which raises confidence, but functional annotation and biological plausibility are not the same as replication. There is no mention of correction for population structure or relatedness within the flock, both easy confounders in a working breeding population. Before using these loci for marker-assisted selection, demand replication in an independent Ordos cohort or a functional follow-up, since suggestive-threshold GWAS hits in small livestock panels have a well-documented tendency not to replicate.
05 Frontiers in Bioinformatics · 19 August 2026_
Human ancestries simulation and inference: a review of ancestral recombination graph-based approaches
- Population genetics
- Review
- Methods
A technical survey of ancestral recombination graph software, weighing the tradeoffs of every major simulation and inference tool built over the last three decades.
Summary
The ancestral recombination graph (ARG) is a central tool in population genetics for representing how recombination and coalescence shape genetic ancestry, but computational cost has long limited its practical use, especially for inference from real data. This review surveys the ARG simulation and inference programs developed over the past three decades, focused on their performance, usability and the biological realism of their underlying algorithms. It is aimed at researchers who want a technical map of the field, including for building their own coalescent-with-recombination methods, and comes with a companion list of software and documentation links.
What to take from it
- Surveys three decades of ARG simulation and inference software
- Evaluates performance, usability and biological realism of each tool's algorithm
- Notes ancestry inference, as opposed to simulation, remains the harder unsolved problem
- Companion resource: compiled links to software, source code and documentation
Review
As a review, its value is in coverage and organisation rather than new results, and coverage is genuinely useful given how fragmented the ARG software landscape has become. The stated three-decade scope, combined with weighing usability alongside algorithmic realism, is a sensible framing for practitioners choosing a tool, but a review is only as current as its writing date - fast-moving ARG inference methods published in just the last two years may already have moved past what is assessed here. It is also written by mathematicians rather than working population geneticists, so cross-check its usability claims against user experience reports before picking a tool based on this alone; the companion software-links resource is worth checking directly for freshness.