01 Proceedings of the National Academy of Sciences · 5 October 2026_
Apomixis imposes evolutionary costs by reducing recombination efficiency and disrupting segregation
- Structural variation
- Population genetics
In wild citrus, a cross between a sexual and an apomictic parent, read against a graph pangenome of both haplotype-resolved parents, shows 24% lower recombination on the apomictic side and segregation distortion over 32% of the genome.
Summary
Citrus hindsii has both sexual and apomictic (adventitious embryony) lineages at the diploid level. The authors crossed a sexual seed parent with an apomictic pollen donor, assembled both parents to haplotype resolution, deep-sequenced 226 F1 progeny and built a graph pangenome to call crossovers. The apomictic lineage carries 1.28 times more heterozygous structural variants, which may suppress crossovers locally, and its recombination efficiency is 24.2% lower than the sexual lineage. Outcrossing through the apomict's pollen exposes recessive load, including a putative gamete-lethal mutation, and 31.7% of the genome shows segregation distortion, driven by several interacting loci.
What to take from it
- Graph pangenome of two phased parents used as the reference for crossover calling in 226 progeny
- Apomictic haplotypes: 1.28x more heterozygous SVs, 24.2% lower recombination
- Segregation distortion over 31.7% of the genome, multi-locus and haplotype-dependent
Review
The design is strong: one cross, phased parents, a pangenome reference, and enough progeny to map crossovers. The weak point is generalization. It is one apomictic and one sexual individual, so the 24.2% figure is a property of these two genomes, not of apomixis; a second pair, or several apomictic lineages, would be needed to separate the mating system from individual history. The link between heterozygous SVs and crossover suppression is described as possible, so check whether crossover deficits actually co-locate with SVs after controlling for local diversity and mappability. 'Recombination efficiency' should be defined precisely - male versus female meiosis differ in rate in many plants, and here the apomict is only the pollen donor. The gamete-lethal mutation is putative. Before using this to design breeding schemes for apomictic crops, ask for replication across crosses, validation of the lethal allele, and a reciprocal cross.
02 Current Biology · 5 October 2026_
Ancient genomes reveal an extinct bovine species from mid-latitude Asia
- Population genetics
- Animal science
Twenty Middle and Late Pleistocene nuclear genomes, many from Denisova Cave, show 'yak X' was a separate species that split from living yaks at least 400,000 years ago and exchanged almost no genes with bison or aurochs.
Summary
Yak X was first seen as a divergent mitochondrial clade in Pleistocene Asia. The authors recovered 20 nuclear genomes from Denisova Cave and other mid-latitude sites, spanning more than 200,000 years. Phylogenetic and population-genomic analyses place yak X as a species distinct from extant yak, diverging at least 400,000 years ago. Its population declined through the Mid-to-Late Pleistocene, with possible turnover and lower diversity around the Last Interglacial (about 130,000-115,000 years ago). Gene flow with steppe bison and aurochs, which lived in the same region and period, was negligible.
What to take from it
- Mitochondrial clade confirmed as a separate species by 20 nuclear genomes
- Split from living yaks at least 400,000 years ago
- No meaningful admixture with bison or aurochs despite co-occurrence
Review
Twenty nuclear genomes across 200,000 years is a large ancient-DNA series for a non-human mammal, and it settles that yak X is not just a mitochondrial oddity. Points to check: genome coverage and damage per sample, since admixture tests on low-coverage ancient genomes have limited power, and 'negligible gene flow' then means 'not detected at this power'. Reference bias matters here: reads mapped to a cattle or yak reference will favour the reference lineage, and the paper should show that D-statistics are robust to the choice of reference. Species status rests on divergence time and lack of admixture, not on morphology or reproductive data, so it is a genomic species call. For livestock genetics, the useful result is that wild Bos and Bison lineages in Pleistocene Asia stayed separate; any claim that yak X contributed to domestic yak or cattle would need direct testing against modern genomes.
03 Nucleic Acids Research · 5 October 2026_
PanGBank: a large-scale resource of precomputed microbial pangenomes built with PPanGGOLiN
- Microbial genomics
- Methods
Precomputed PPanGGOLiN pangenome graphs for over 4,600 GTDB species and 393,000 genomes, with persistent/shell/cloud partitions, genomic plasticity regions and modules, behind a REST API, CLI and web interface.
Summary
PanGBank provides species-level prokaryotic pangenomes built with PPanGGOLiN partitioned graphs. The first release has two collections: GTDB_all, which includes metagenome-assembled and single-cell genomes for maximum diversity, and GTDB_refseq, restricted to high-quality annotated genomes. Each pangenome has gene families partitioned into persistent, shell and cloud, regions of genomic plasticity (panRGP) and co-localized functional modules (panModule). A use case traces antimicrobial resistance determinants across the Acinetobacter baumannii pangenome.
What to take from it
- Over 4,600 species, 393,000 genomes, built with one pipeline
- Two collections: all GTDB genomes including MAGs/SAGs, or RefSeq-quality only
- Access by REST API, command line and web
Review
A uniform, versioned resource removes the main source of irreproducibility in bacterial pangenome work: every lab running its own clustering with its own thresholds. The cost is that every result inherits PPanGGOLiN's choices - its clustering identity and coverage cut-offs and its statistical partition model - so users should check those parameters before comparing PanGBank partitions to results from Panaroo or Roary. The GTDB_all collection includes MAGs and SAGs, which are often incomplete and contaminated; incompleteness pushes genes from persistent to shell or cloud, and contamination adds false cloud genes. Use GTDB_refseq for any partition-level claim, or check that conclusions hold in both. Ask how often the resource will be rebuilt as GTDB releases change species boundaries, and whether old versions stay accessible for reproducibility.
04 Microbiology and Molecular Biology Reviews · 5 October 2026_
The grammar of the pangenome: terms, concepts, methods, and emerging patterns - a general framework for comparative genomics and genome biology
- Review
- Microbial genomics
- Methods
A review of pangenome terms, open/closed models and software, with a re-analysis of 249 taxa and 17,858 genomes and proposed reporting criteria and standard operating procedures.
Summary
The review covers estimation of gene and protein family numbers, models of open and closed pangenomes, and definitions needed for reporting guidelines. It lists and compares computational platforms, algorithms and data structures. From a validated set of 249 taxa and 17,858 genomes it reports patterns such as novel families per added genome and pangenome size estimates. It proposes revised criteria for what counts as a pangenome, standard operating procedures, and targets for reproducibility.
What to take from it
- Definitions and reporting criteria for pangenome studies
- Comparison of pangenome software and data structures
- Pangenome size and novelty estimates from 249 taxa, 17,858 genomes
Review
Useful as a reference for terms and as a checklist of what a pangenome paper should report. Its scope is gene-content pangenomes, mostly prokaryotic; sequence-graph pangenomes of the kind used in human, plant and livestock genomics (minigraph, PGGB, vg) are a different object, and readers from that side should not expect guidance on graph construction, variant calling from graphs or reference bias. The 249-taxon patterns depend on the clustering and sampling used, so pangenome size and openness estimates from it are comparable only to analyses run the same way. The proposed SOPs are valuable only if journals or tools adopt them; check whether they come with a minimal reporting table that can be filled in, rather than prose recommendations.
05 PLOS ONE · 5 October 2026_
Sub-consensus haploid variant calling in long-read sequencing technology
- Methods
- Microbial genomics
LoFreq finds variants down to 0.1% frequency in Oxford Nanopore reads but with a chemistry- and depth-dependent false discovery rate; a Phred-score recalibration cuts false positives above about 10% frequency.
Summary
LoFreq was built for short reads. The authors benchmarked it on ONT data using truth sets of three SARS-CoV-2 spike plasmids (7,179 bases, 100 SNVs) and whole Escherichia coli genomes, sequenced on R9.4.1 and R10.4.1 flow cells. Sensitivity was high, down to 0.1% allele frequency, and best on R10.4.1, but false discovery rates varied with chemistry and depth. A Phred quality recalibration reduced false positives while keeping recall, though it helped less below 10% frequency and does not apply to structural variants.
What to take from it
- LoFreq on ONT reads: high recall, uncontrolled false discovery rate
- R10.4.1 performs better than R9.4.1
- Recalibration helps above about 10% frequency; not for SVs
Review
A practical benchmark with known truth sets, but narrow. Plasmid mixtures and E. coli do not test homopolymers and low-complexity regions at the scale of a eukaryotic genome, or the amplicon dropouts of real viral surveillance data. The recalibration is fitted to these runs; ask whether it transfers to a new basecaller version, since ONT error profiles change with each release and the correction may need refitting every time. The method leaves the hardest range - below 10%, where within-host variants usually sit - largely unsolved. It should also be compared with callers built for long reads, such as Clair3 or other ONT-aware low-frequency callers, before anyone chooses LoFreq for this job.