01 Nature Genetics · 25 September 2026_
Genome-wide mapping of common and rare variant effects on adiposity across childhood
- Human genetics
Age-stratified GWAS of measured (n = 62,276) and recalled (neff = 599,924) childhood adiposity finds 624 loci, a third with no matching effect on adult BMI, plus rare coding variants in MC4R, POMC and others that act more strongly in childhood.
Summary
The authors ran genetic association studies stratified by age on objectively measured childhood adiposity traits and on recalled childhood body size. They report 624 common variants; about one third show no concordant association with adult BMI. Loci in the leptin-melanocortin pathway (BSX, GNAS, LEPR, PCSK1) and incretin signalling (GIPR, GLP1R) have childhood-specific effects. Single-nucleus RNA-seq points to cell populations in the arcuate nucleus and mammillary bodies. Exome sequencing in 479,615 people finds rare protein-coding variants in ADCY3, CALCR, MC4R, MRAP2, POMC and MYH13 with stronger effects in childhood than in adulthood.
What to take from it
- 624 childhood adiposity loci; one third not associated with adult BMI
- Leptin-melanocortin and incretin loci act specifically in childhood
- Rare coding variants in MC4R, POMC, ADCY3 and others: larger effects in children
- Design point: stratifying by age separates effects that pooled adult GWAS average out
Review
The design lesson transfers directly: effects that change with developmental stage are diluted when phenotypes from different ages are pooled, which applies equally to growth traits in livestock. The weak point is the phenotype that drives most of the power: recalled childhood body size is a coarse, self-reported category, and 'no concordant adult association' for a third of loci could partly reflect differences in measurement and power between the childhood and adult analyses rather than true age specificity. The abstract does not state ancestry composition. Before using these loci, check how many replicate in the measured-adiposity subset alone, and whether the childhood-specific calls survive a formal test of effect-size heterogeneity rather than a difference in significance.
02 Nature · 23 September 2026_
The mutational dynamics of the Arabidopsis centromeres
- Structural variation
- Population genetics
Replicated assemblies of Arabidopsis mutation-accumulation lines show centromeric point mutations arise at almost ten times the genome-wide rate, and kilobase indels add or remove only whole repeat units.
Summary
Using replicated genome assemblies of mutation-accumulation lines, the authors measured mutations inside centromeric tandem-repeat arrays directly. Kilobase-sized indels are frequent and always change the array by complete repeat units, so the array structure is kept. Point mutations accumulate nearly tenfold faster than elsewhere in the genome, probably through non-allelic gene conversion between nearby repeat units. Lines lacking the anti-recombinase helicase RTEL1 have more and longer repeat-preserving indels, which implicates homology-directed repair. Forward simulations using the measured spectrum reproduce megabase-scale homogenized repeat blocks like those in natural centromeres.
What to take from it
- First direct centromere-specific mutation spectrum from mutation-accumulation lines
- Point mutation rate in centromeres about 10x the genome average
- Kb indels change arrays only by whole repeat units
- RTEL1 loss increases indel frequency and length: homology-directed repair drives it
- Simulation with this spectrum alone produces natural-looking homogenized arrays
Review
This gives an empirical mutation model for tandem arrays, which pangenome graphs and SV callers currently handle poorly: centromeric differences between assemblies are expected to be unit-sized indels plus gene-conversion SNPs, not random. The main risk in any study of mutations inside satellite arrays is assembly error being counted as mutation; the replicated assemblies are the authors' control, so the replication concordance and validation rate in the full paper are what to check first. The result comes from one selfing species over a limited number of generations, so rates may differ in outcrossing plants or in animals with different centromere repeat types. Demand the per-generation rate with confidence intervals before using it as a prior in population models.
03 Nature Communications · 25 August 2026_
Causal inference for multiple risk factors and diseases from genomics data
- Methods
- Human genetics
Causal Inference GWAS picks genetic instruments from summary statistics while controlling for linkage and pleiotropy; across 9 risk factors and 4 metabolic diseases in UK Biobank it keeps only 696 valid instruments and finds most known causal paths indistinguishable from confounding.
Summary
Mendelian randomization is biased by poor instrument selection and horizontal pleiotropy. The authors propose a graphical method that selects instruments from summary statistics while accounting for linkage and pleiotropy, handles rare variants and binary outcomes, separates direct from indirect risk factors, and flags latent confounding. Applied to 9 risk factors, 4 metabolic disease outcomes and 8.4 million variants in 458,747 UK Biobank participants, it runs in 20 minutes and retains 696 genome-wide valid instruments. It replicates almost all previously reported risk factor-outcome paths but finds most of them indistinguishable from unmeasured confounding.
What to take from it
- Instrument selection from summary statistics, controlling for LD and pleiotropy
- Separates direct from indirect risk factors; flags latent confounding
- UK Biobank: 696 valid instruments across 9 risk factors and 4 diseases
- Most published risk factor-outcome paths cannot be separated from confounding
- Authors argue longitudinal and family data are needed for causal claims
Review
The useful finding is negative: when instruments are filtered strictly, most MR-style causal claims in biobank data do not survive. That conclusion depends on the assumptions of the graphical model, such as faithfulness and correct handling of LD, which are hard to verify in real data. Check the simulations for calibration under realistic pleiotropy and sample overlap, and compare against current robust MR estimators on the same data. The analysis is UK Biobank only, largely European ancestry. For livestock work the approach could screen genetic correlations between traits for direct effects, but pedigree structure and much smaller samples would need testing first.
04 Biology · 16 September 2026_
Genomic Diversity and Footprints of Selection in Hamdani Sheep
- Animal science
- Population genetics
ddRADseq of 18 Hamdani sheep (310,663 SNPs) finds high diversity, low inbreeding, a cluster distinct from Turkish fat-tailed breeds, and preliminary selection candidates near TSHR, FSHR and LHCGR.
Summary
Twenty Hamdani sheep were genotyped by double-digest RAD sequencing; 18 animals and 310,663 SNPs passed filtering. Observed and expected heterozygosity were 0.323 and 0.326; FHOM was 0.029 and FROH 0.008. Population structure separates Hamdani from Turkish fat-tailed breeds. ROH and iHS scans give candidate regions; after FDR correction, 12 of 34 iHS trait-chromosome combinations and 3 of 8 ROH combinations were enriched. Candidate genes include reproduction genes (TSHR, FSHR, LHCGR), lipid metabolism genes (ABCG1, LPIN1) and adaptation genes (LACC1, MBOAT7).
What to take from it
- 18 animals, 310,663 ddRAD SNPs
- Ho 0.323, He 0.326; FROH 0.008
- Distinct from Turkish fat-tailed breeds
- Candidates near TSHR, FSHR, LHCGR, ABCG1, LPIN1 - labelled preliminary by the authors
Review
Useful as a first diversity baseline for an under-studied breed, not as evidence of selection. Eighteen animals is too few for haplotype statistics: iHS needs phased haplotypes and reasonable sample size, and ROH calling from sparse, unevenly spaced ddRAD markers underestimates short ROH, which may explain the very low FROH. The candidate genes are ones that appear in most sheep selection scans, so their recurrence here adds little. Before using any region, demand whole-genome resequencing or a denser array, a larger sample, and a comparison with FST or XP-EHH against named reference breeds.
05 International Journal of Molecular Sciences · 8 September 2026_
Genetic Diversity and Selection Signatures of Two North American Corriente Cattle Herds in Northeastern New Mexico, United States
- Animal science
- Population genetics
A 64K SNP array on 52 North American Corriente cattle from two New Mexico herds suggests a distinct Criollo lineage and selection candidates overlapping QTL for movement, fertility, immunity and carcass traits.
Summary
North American Corriente is a Criollo-derived breed used for rodeo and beef. Of 97 animals genotyped on a ~64K SNP array, 52 remained after quality control, relatedness filtering and structure analysis. Admixture and structure analyses indicate some differentiation consistent with a distinct Criollo lineage, which the authors call preliminary. Selection-signature regions overlap known QTL for movement, fertility, immune response, and carcass and milk traits.
What to take from it
- 52 animals from two commercial herds, ~64K SNP array
- Genetic differentiation consistent with a distinct Criollo lineage
- Candidate regions overlap QTL for movement, fertility, immunity, carcass and milk
- Authors state conclusions are preliminary
Review
The study documents a breed with almost no genomic data, and that is its value. Two herds from one region means herd structure and breed structure cannot be separated, and the retained 52 animals came after removing close relatives, which leaves little power for selection scans. Commercial SNP arrays are designed from mainstream taurine breeds, so ascertainment bias can hide Criollo-specific diversity. Overlap with cattle QTLdb is weak evidence because the database covers much of the genome. Before relying on it, demand samples from independent herds across the breed's range, sequence data or ascertainment correction, and comparison with other Criollo and Iberian reference populations.