The framework consolidates checks that would otherwise need to be conducted separately for each EDTA result directory. These include the presence of expected output files, reported transposable-element content and the sizes of transposable-element-library files. EDTA-Audit is a post-processing tool rather than a new annotation method: it examines completed EDTA outputs and summarizes whether they satisfy predefined checks.

Its batch reports provide an organized first-pass screen for prioritizing results that require further attention, without replacing expert review. Evaluation combined a pilot analysis with a full-dataset summary. Thirteen of the 142 assemblies were selected as a pilot case set to examine the behavior of the default Plant Mode thresholds.

Audit outcomes were then summarized across all 142 assemblies. During this evaluation, EDTA-Audit generated PASS or FAIL labels and diagnostic explanations for flagged outputs. Reported functions included detecting missing required files and highlighting atypical transposable-element-content or transposable-element-library profiles.

The results position EDTA-Audit as a practical quality-control utility for batches of plant genome assemblies. By combining file checks with taxon-informed profile warnings, the software gives users a consistent way to separate outputs that meet the configured criteria from those that need closer inspection. The reported evaluation was limited to the tested dataset, default Plant Mode settings and version 1.0.0, and the tool did not establish that a PASS label guarantees the biological accuracy of an annotation.

The authors instead describe its role as a first-pass screen to organize review and focus troubleshooting. EDTA-Audit v1.0.0, the version used to generate all reported results, was implemented in Bash. The repository has since been updated to a Python-based v2.0.0 implementation, developed with the aim of improving usability, maintainability and extensibility.

This later implementation does not alter the scope of the reported evaluation, because every result presented in the study came from v1.0.0. The study was published as accepted research in Plant Methods ; the supplied early version carries a permanent DOI but remains subject to further editorial changes before replacement by the final version of record.

Subject of Research: Quality control of EDTA-based transposable-element annotations in plant genome assemblies

Article Title: EDTA-Audit: a taxonomy-aware post-processing framework for quality control of EDTA-based transposable element annotation

Article References: Liu, D., He, S., & Zhang, Y. (2026). EDTA-Audit: a taxonomy-aware post-processing framework for quality control of EDTA-based transposable element annotation. Plant Methods. https://doi.org/10.1186/s13007-026-01607-5

Image Credits: AI Generated

DOI: 10.1186/s13007-026-01607-5

Keywords: EDTA-Audit, transposable elements, plant genomes, bioinformatics, quality control, genome annotation, batch processing, taxonomy-aware screening, software, taxonomy-aware, post-processing, framework

The Chaetothyriales are a fascinating order of ascomycete fungi, characterized by their dark, melanin-rich cell walls and their remarkable ability to colonize extreme environments. Ancestral state reconstructions suggest the order originated from melanized, slow-growing rock-inhabiting fungi or lichen-associated ancestors, and today its members occupy a striking diversity of niches: some live on bare rock exposed to punishing ultraviolet radiation, others associate with ants, lichens, or plants, and a few have evolved as opportunistic pathogens of humans. The order is currently divided into at least seven families, including the Herpotrichiellaceae, which contains thermophilic human pathogens such as Cladophialophora carrionii, the agent of chromoblastomycosis. The Epibryaceae, first formally described in 2014, sit near the base of this evolutionary tree, making them a key lineage for understanding how the ecological diversity of the entire order came to be.

Until now, genomic information on the Epibryaceae was almost entirely absent, which hindered any serious attempt to understand the molecular basis of their apparent preference for cold. The genus Epibryon comprises roughly fifty described species, most of them tiny fungi that form small, dark, bristly fruiting bodies measuring just 40 to 150 micrometers across. They typically grow on the older, dying, or dead leaves of mosses and leafy liverworts, sometimes showing remarkable specificity for particular host species or even particular microsites such as leaf nerves or individual cells. A few species have been reported from rock surfaces, soil, or as epiphytes, and several were described from high-latitude regions, hinting at a link between the lineage and low temperatures long before genomic tools could test that idea.

To close this gap, the researchers selected six reference strains spanning the diversity of the family: Epibryon bryophilum, E. humicola, E. interlamellare, E. minutissimum, E. sylvestris, and E. turfosorum. The fungi were cultured on Sabouraud’s Glucose Agar at 28 degrees Celsius for a week, and DNA was extracted using the classic CTAB method with chloroform and isoamyl alcohol. Sequencing was carried out on Illumina platforms, including the NovaSeq 6000, MiSeq, and BGI-Seq500, generating 150-base-pair paired-end reads. The team then assembled the genomes de novo using SPAdes with multiple k-mer sizes, filtered out non-eukaryotic contamination with the deep-learning tool Tiara, and assessed completeness with BUSCO against a Chaetothyriales-specific database. Gene prediction and annotation were performed with the Funannotate pipeline, drawing on databases such as InterPro, PFAM, dbCAN, MEROPS, UniProt, and EggNOG to assign functions to the predicted proteins.

The resulting genomes averaged 29.09 megabase pairs in size, ranging from 26.20 to 34.97 Mbp, and contained an average of 10,495 predicted genes. BUSCO completeness averaged 85.8 percent using a set of 6,265 conserved fungal genes, a respectable figure for these fragmented assemblies. The researchers then compared these new genomes with publicly available references across the subclass Chaetothyriomycetidae, including 62 genomes from the Chaetothyriales itself, plus representatives of the Phaeomoniellales, Pyrenulales, and Verrucariales. Orthology analysis with OrthoFinder2 classified genes into orthogroups, and a phylogenomic tree built from 296 single-copy orthogroups confirmed that all Epibryon species share a common ancestor and that the family forms a distinct, early-diverging branch within the order.

One of the most striking taxonomic outcomes concerns Cladophialophora brunneola, a fungus isolated from karst rock at 2,220 meters altitude in Guizhou, China. Previously described as an early-diverging lineage within the Chaetothyriales, this species clearly clusters within the Epibryaceae in both the phylogenomic and multilocus analyses, which were based on the ribosomal genes SSU, ITS, and LSU together with the protein-coding RPB2 gene. The authors therefore formally reclassified it as Epibryon brunneolum comb. nov. Interestingly, this rock-dwelling species showed the highest proportion of orthogroups containing two genes, suggesting greater protein diversity, and it also possesses the highest number of carbohydrate-active enzymes in the family—features that may reflect the severe fluctuations in temperature and humidity experienced by fungi living on sun-exposed rock surfaces.

The functional comparisons revealed a distinctive enzymatic toolkit that sets the Epibryaceae apart from their relatives. On average, family members carry 297 genes assigned to the CAZyme database, spanning 106 families, and 345.7 genes in the MEROPS peptidase database, spanning 89 families. Crucially, the Epibryaceae show a reduced set of enzymes associated with lignin degradation, including contractions in the auxiliary activity families AA7 and AA9 and the glycoside hydrolase family GH12. They also uniquely possess the carbohydrate-binding module CBM52, which binds to endo-1,3-beta-glucanase and enhances its activity during cell septation, though its precise ecological role remains to be elucidated. On the peptidase side, the family uniquely harbors the enzyme U74, which has a nonspecific catalytic site capable of digesting proteins without any restriction on substrate size, and shows expansions in the S53 and T02 families alongside contractions in C26 and C85.

The most compelling evidence for cold adaptation comes from the analysis of stress-response genes. All Epibryaceae species possess the cold shock protein CspA, an RNA-binding enzyme that stabilizes ribosomal structures and maintains protein translation under cold stress. No other family in the Chaetothyriales carries this protein, although it is known to be expanded among black fungi isolated from Antarctic environments in the class Dothideomycetes, underscoring its importance for fungal survival at low temperatures. At the same time, the Epibryaceae lack the ClpA/B conserved site 2, a domain associated with protecting cellular structures during heat stress, suggesting a reduced tolerance for high temperatures. The family also shows a duplication of 5-formyltetrahydrofolate cyclo-ligase in its last common ancestor and an expansion of the DNA recombination and repair protein Rad51, which has five copies compared with four in the ancestral Chaetothyriales and only two in several Herpotrichiellaceae. Rad51 repairs DNA damage caused by ultraviolet light, and its expansion may help these fungi withstand the intense radiation of glacial and high-altitude environments.

These genomic signatures align neatly with the ecological record. Mapping the isolation sites of Epibryaceae strains onto the Köppen climate classification shows that most known isolates come from polar regions and cold or temperate continental zones, with precipitation distributed throughout the year. The family’s reduced capacity for lignin and complex polysaccharide breakdown fits its saprobic lifestyle on mosses and liverworts, which offer readily accessible carbon sources in nutrient-poor, cold environments where lignin-rich plant material is scarce. The authors suggest that this combination of traits—a psychrophilic preference, a saprobic ecology on bryophytes, and occasional associations with lichens, soil, and the phyllosphere—distinguishes the Epibryaceae from all other chaetothyrialean lineages and may reflect an ancestral ecological state shared with closely related orders such as the Verrucariales, many of which are lichen-associated.

The study also highlights how much remains to be discovered. Despite extensive herbarium records, molecularly characterized isolates of Epibryon remain scarce, and the true breadth of the family’s ecological distribution is only beginning to emerge. The finding that a genus once defined by its moss-dwelling habit also includes rock-inhabiting and soil-dwelling species suggests that future sampling from underexplored cold regions will likely reveal further diversity and refine our understanding of its specialization. For now, the genomes of these six strains provide the first molecular window into a lineage that traded heat tolerance for cold resilience, and they demonstrate how comparative genomics can transform a little-known family of moss-inhabiting microfungi into a model for studying fungal adaptation to some of the planet’s most demanding environments.

Subject of Research: Genomics and cold-adaptation mechanisms of the psychrophilic fungal family Epibryaceae in the order Chaetothyriales

Article Title: Genomics and ecology of Epibryaceae , a psychrophilic family in Chaetothyriales

Article References: Lustosa, B. P. R., Belmonte-Lopes, R., de Hoog, S., Costa, F. D. F., Jacomel, B., dos Santos, G. D., Razzolini, E., Li, Y., Xue, R., Baura, V. A., Souza, E. M. D., Gomes, R. R., Ahmed, S. A., Selbmann, L., Song, Y., & Vicente, V. A. (2025). Genomics and ecology of Epibryaceae, a psychrophilic family in Chaetothyriales. IMA Fungus, 16, Article e170120. https://doi.org/10.3897/imafungus.16.170120

Image Credits: AI Generated

DOI: 10.3897/imafungus.16.170120

Keywords: Epibryaceae, Epibryon, Chaetothyriales, psychrophilic fungi, comparative genomics, cold shock protein CspA, CAZymes, MEROPS peptidases, phylogenomics, bryophytes, black fungi, genome annotation

The research team, led by Wenkai Duan and Lingyao Peng, assembled volatile, flavonoid-feature, amino-acid and electronic-sensing profiles across nine commercial dried-fruit groups: seven sampling groups spanning six Zhejiang county-level areas plus two Anhui counties, including both cultivated and wild material from Pan’an. Zhejiang material is traded as a representative authentic (daodi) medicinal product within the Zhe Ba Wei group, making it a natural test case for chemical authenticity. But the compiled chemical data were not generated de novo for the study; they were aggregated as location-level summaries from vendor and laboratory reports, with historical sampling records referring to three biological replicates per group whose replicate-level observations were not preserved. The authors are explicit about these metadata limits: plant-level sampling, harvest date, genotype, maturity criteria and drying history cannot be resolved from the available records.

Four analytical platforms contributed to the chemical landscape. Headspace solid-phase microextraction gas chromatography–mass spectrometry profiled 124 volatile feature rows, all treated as library- and retention-index-supported putative annotations because raw spectra and a complete match ledger were unavailable. Nine laboratory reports quantified 17 amino acids in submitted dried material, with tryptophan reported as not detected in all nine — a method-limited result, since prolonged acid hydrolysis compromises tryptophan recovery. A targeted HPLC-UV method captured 12 characteristic flavonoid peaks, while an LC-MS/MS dataset on a QTRAP platform yielded 318 location-level feature rows, again treated as putative unless authentic-standard support was explicit. Electronic-nose and electronic-tongue measurements added holistic fingerprints, but the authors treated electronic-nose data as technical repeats only and retained electronic-tongue vendor outputs as report-level records rather than reanalysing them as independent geographic evidence.

The descriptive chemistry is striking. Benzaldehyde was highest in Shengxian (F4) and lowest in Pan’an cultivated (F8); linalool was highest in Yuanqiao (F2) and lowest in Bozhou (F1). The sum of quantified amino acids was highest in Yuanqiao at 83,537.9 mg/kg and lowest in Pan’an wild material at 63,751.2 mg/kg, with glutamate and aspartate pools dominating — interpreted as combined Glu+Gln and Asp+Asn because acid hydrolysis converts glutamine and asparagine. Putative rutin was highest in Pan’an cultivated and lowest in Pan’an wild. But the authors refuse to treat any of these rankings as marker evidence. Instead, they introduce a prespecified attribution gate: a feature could be prioritised only when its observed nine-location range exceeded the developmental range measured in an independent four-stage dataset.

That independent dataset came from a public NCBI project (PRJNA671545) containing 12 paired-end fruit RNA-seq libraries — three biological replicates at each of four publication-defined stages: big green, green to yellow, yellow to orange and red — plus a developmental metabolite table with 24 samples and 21 features. For each of 11 name- or formula-level overlaps between the location and developmental datasets, the team compared the nine-location log2 range (Rloc) with the four-stage developmental log2 range (Rdev). Developmental-envelope coverage was defined as C = min(Rdev/Rloc, 1), and residual range excess as E = max(Rloc − Rdev, 0). Positive E values identify candidates whose location range exceeds the observed developmental envelope. Twenty of 21 public metabolites showed a stage effect after Benjamini-Hochberg correction, and all 11 overlaps were stage-sensitive. Eight overlaps had developmental ranges equal to or exceeding location ranges, giving C = 1 and no residual excess.

Three candidates retained positive E: gallic acid (Rloc 3.40, Rdev 3.00, E 0.41 log2 units), kaempferol 3-O-glucoside (1.30, 0.92, E 0.38) and ellagic acid (1.49, 0.57, E 0.92). Robustness was assessed by deleting each location once and recalculating E — gallic acid retained positive E in 8 of 9 leave-one-location-out analyses, while ellagic acid and kaempferol 3-O-glucoside did so in 9 of 9. Scenario sensitivity was then evaluated with 30,000 Monte Carlo runs per setting, each generating nine synthetic lots from Dirichlet stage mixtures (α = 0.5, 1 or 2), bootstrapping the six public observations per stage and applying an assumed lot-level drying factor on the log2 scale (SD = 0, 0.25 or 0.5). Tier 1 required positive full-data E, retention in at least 8 of 9 leave-one-location-out analyses and an observed range above the simulated 95th percentile in all three settings at moderate drying SD. Gallic acid and ellagic acid met Tier 1; kaempferol 3-O-glucoside exceeded the scenario threshold in only one of three settings and was assigned Tier 2. With an assumed drying SD of 0.5, no candidate was robust across all mixture settings — a sobering quantification of how much unmeasured commercial-lot structure can matter.

The second evidence gate turns to the genome. The chromosome-scale R. chingii reference genome (v1.0) had reported terpene-synthase activity among expanded gene functions but never provided a sequence-resolved terpene-synthase catalogue. The team scanned all 33,130 predicted proteins with the PF00201 UDP-glycosyltransferase profile using PyHMMER, identifying 178 protein models. Only 55 of these (30.9%) carried a UGT-related KEGG orthology label, one carried a discordant non-UGT label and 122 had no KO assignment at all. The family-profile catalogue recovered all 172 published RchUGT loci — an expected consistency check, since that catalogue was itself built by a family-profile search of the same genome — plus six additional profile-defined candidates that may include fragments or split gene models pending transcript-supported curation.

A species-matched positive-control benchmark quantified the sensitivity gained over KO-only screening. Three publicly deposited and experimentally cloned R. chingii UGT coding sequences — RcUGT75L20, RcUGT75T4 and RcUGT85A57 — were used as controls. The PF00201 scan recovered all three, whereas the prespecified KO-only flavonol-glycosyltransferase screen and the PF00201-plus-UGT-related-KO intersection each compiled none. Across 10,000 random three-protein sets sampled from the proteome, the mean number of PF00201 hits was 0.0171 and no set had three; the exact hypergeometric probability was 1.53 × 10⁻⁷. The authors caution that this benchmark largely restates how few proteins the profile hits, and should be read as a sensitivity check rather than specificity evidence. On the terpene-synthase axis, Pfam screening detected 67 protein models containing at least one TPS domain, of which 32 combined a C-terminal domain with a class-I or class-II catalytic motif. These motif-supported candidates were non-randomly distributed: LG03 contained 17 of 32, a 3.3-fold enrichment over expectation (hypergeometric P = 1.52 × 10⁻⁶), with a median adjacent physical distance of 54.8 kb compared with 1.41 Mb across 10,000 random 17-gene sets.

The most specific genome result concerns one focal gene model. LG03.2235 encoded a 1234-amino-acid model with both TPS domains and catalytic motifs but carried an MRE11 DNA-repair KO — an annotation conflict. Transcript evidence showed ordinary junction support for 21 of 26 annotated junctions, yet the 22,832-bp composite boundary received zero exact junction reads in all 12 fruit libraries. Comparative genomics then resolved the concern: in red raspberry (Rubus idaeus cultivar Varnes), the LG03.2235 N-terminal and TPS segments occupied two non-overlapping, same-strand homologous blocks separated by 4.5 kb; in black raspberry (Rubus occidentalis v3.0), the same segments mapped at 95.2% and 90.5% identity to two separate proteins 46.7 kb apart. By contrast, LG02.1599 — an unusually long ent-copalyl diphosphate synthase model — mapped to one continuous homologous interval in both near relatives, supplying an internal negative comparator. The agreement among annotation conflict, transcript non-support and cross-species separation supports a specific fused-model annotation concern requiring re-curation, though the corrected transcript must still be established by transcript long reads or targeted splice-aware curation.

The study’s practical output is a two-gate prioritisation framework rather than a validated origin classifier. No geographic classifier was fitted to the location-summary matrices, and the authors specify the sample matrix, grouped resampling, held-out predictions and calibration outputs that a reproducible future evaluation would require. Environment-metabolite associations tested at the location level (n = 9) found no reproducible signal: altitude was not associated with linalool (permutation P = 0.409) or rutin (P = 0.882), and sunshine hours were not associated with linalool (P = 0.880), with all adjusted q values at 0.882. With nine independent locations, approximate 80% power is limited to correlations near |ρ| = 0.84, so moderate associations remain unresolved — and province and altitude are structurally entangled because both Anhui locations are low altitude. The four-stage expression-metabolite coupling audit across 247 unique candidate genes and 21 metabolites (3,087 pairs) nominated trajectories but established no regulation: the minimum attainable exact two-sided P was 0.0833, no pair survived correction and the minimum q was 0.411. The framework’s value lies in converting two common uncertainties — missing maturity metadata and incomplete functional labels — into auditable validation priorities. Gallic acid and ellagic acid begin targeted standard confirmation and stage-controlled commercial-lot replication first, followed by kaempferol 3-O-glucoside, while the remaining eight overlaps are retained as development-sensitive quality descriptors rather than discarded. The workflow can transfer to other processed fruits in which harvest stage is missing or incompletely recorded, provided candidates first pass authentic-standard confirmation and moisture-aware quantification, then independent lots with recorded maturity, drying and storage. Only after those gates should location-grouped classification or locus-specific expression be considered — and any such classifier must fit every preprocessing and feature-selection step inside location-grouped training folds, reporting per-location held-out predictions, calibration and uncertainty.

Subject of Research: Food Chemistry

Article Title: Developmental-envelope screening and genome-annotation audit prioritize chemical authenticity candidates in commercial Rubus chingii fruit

Article References: Duan, W., Wei, Z., Gao, W., & Peng, L. (2026). Developmental-envelope screening and genome-annotation audit prioritize chemical authenticity candidates in commercial Rubus chingii fruit. Food Chemistry: X, 40, Article 104635. https://doi.org/10.1016/j.fochx.2026.104635

Image Credits: AI Generated

DOI: 10.1016/j.fochx.2026.104635

Keywords: food authenticity, Rubus chingii, developmental envelope, genome annotation, UGT, TPS, fupenzi, metabolomics, Developmental-envelope, screening, genome-annotation, audit

The newly assembled mitochondrial genome of P. tuberculosus measures 115,079 base pairs with a GC content of 26.1 percent, a size and composition that fit comfortably within the range reported for closely related Phellinus species. For fungi, whose mitochondrial genomes can vary enormously in size and architectural complexity, this figure immediately signals both conservation and room for the kind of expansion that makes fungal mitogenomics so intriguing. Unlike the compact animal mitochondria most people learn about in school, fungal mitochondrial genomes routinely swell with introns, repeats and intergenic spacers, and the Phellinus genomes are no exception.

Annotation of the P. tuberculosus mitogenome revealed a core genetic toolkit that appears to be a hallmark of the group: fourteen core protein-coding genes, twenty-seven transfer RNA genes and two ribosomal RNA genes. Those fourteen protein-coding genes correspond to the standard set required for oxidative phosphorylation, the electron transport chain machinery that mitochondria use to generate cellular energy. Their consistent presence across the compared species underscores how tightly evolution preserves the essential respiratory apparatus, even as the surrounding genomic landscape churns with rearrangements, duplications and mobile elements.

Beyond the genes themselves, the researchers catalogued 79 simple sequence repeats along with multiple larger repeat sequences scattered across the genome. Simple sequence repeats, sometimes called microsatellites, are short tandemly repeated motifs that mutate rapidly and can serve as informative markers for population-level studies. Larger repeat elements, by contrast, are the raw material of genome rearrangement: because recombination between repeats can invert, shuffle or duplicate genomic segments, their abundance helps explain why fungal mitochondrial genomes can differ dramatically in gene order even between closely related species. The repeat landscape of P. tuberculosus thus provides both a record of past genomic turbulence and a potential engine of future change.

One of the most technically interesting findings concerns RNA editing, the post-transcriptional process that alters the sequence of RNA transcripts relative to the DNA template. The team predicted a total of 42 C-to-U RNA editing sites across the mitochondrial protein-coding transcripts, with the nad2 gene carrying the highest number of editing sites. C-to-U editing in fungal mitochondria changes cytidine residues to uridine in the RNA, sometimes correcting codons so that the final proteins match functional expectations even when the gene sequence alone would suggest otherwise. The concentration of editing sites in nad2, which encodes a subunit of the respiratory chain’s NADH dehydrogenase complex, hints that this gene in particular may rely on post-transcriptional correction to produce a properly functioning protein.

Comparative genomic analysis with the two previously characterized Phellinus mitochondrial genomes showed that the P. tuberculosus lineage shares conserved mitochondrial genome features with its relatives, from gene content to overall organization. At the same time, the study’s title points to dynamic genome evolution within Hymenochaetaceae more broadly, reflecting the tension that defines fungal mitogenomics: the core is stable, but the periphery is restless. Introns, in particular, are notorious agents of mitochondrial genome expansion in basidiomycetes, and their gain and loss across lineages can inflate genome size without adding protein-coding capacity. The Phellinus comparison captures this dynamic in miniature, showing how genomes that carry the same essential genes can nonetheless diverge substantially in size and structure.

The phylogenomic component of the study used mitochondrial sequence data to test how P. tuberculosus relates to other members of its genus and family. The analyses showed that P. tuberculosus clusters with the two previously characterized Phellinus species in a strongly supported clade, a result that both validates the mitochondrial genome as a phylogenetic marker and reinforces the monophyly of the genus as currently sampled. Strong statistical support for this grouping matters because Phellinus, like many wood-inhabiting polypore lineages, has a taxonomic history complicated by morphological similarity, and molecular data have repeatedly reshaped how these fungi are classified. A well-supported mitochondrial phylogeny offers an independent line of evidence for species boundaries and evolutionary relationships.

Why does a mitochondrial genome from a single mushroom species matter beyond taxonomy? Phellinus species are white-rot fungi, meaning they possess the enzymatic machinery to degrade lignin, the tough aromatic polymer that gives wood its rigidity. That capability has enormous ecological significance in forests, where white-rot fungi drive carbon cycling by decomposing dead wood, and equally significant biotechnological potential, since lignin-degrading enzymes are of interest for biofuels, bioremediation and green chemistry. Many Phellinus species also produce medicinally functional compounds and nutrients, and several have been used in traditional medicine systems for centuries. Understanding the genetic architecture of these organisms, including their mitochondrial genomes, contributes to a fuller picture of their biology, their evolution and their potential applications.

The authors frame their work as filling a genuine gap: despite the medicinal importance of Phellinus, a comprehensive genus-level analysis of mitochondrial genomes had been unavailable until now. By providing the first complete mitogenome for P. tuberculosus and comparing it with its congeners, the study establishes a baseline for future phylogenetic, ecological and biotechnological investigations of medicinally important species. The annotated genome, with its catalogued repeats and predicted RNA editing sites, becomes a resource that other researchers can mine for markers, for insights into mitochondrial gene expression, and for reconstructing the evolutionary history of a lineage that shapes forests and pharmacies alike.

The study also illustrates a broader trend in genomics: mitochondrial genomes, once treated as an afterthought compared to nuclear genomes, have become workhorses for evolutionary biology in fungi. Their relatively small size makes them cheap to assemble completely, their uniparental inheritance simplifies comparisons between lineages, and their combination of conserved cores and variable architecture makes them ideal for resolving relationships at shallow and intermediate evolutionary depths. For the Hymenochaetaceae, a family that includes numerous pathogens and decomposers of standing trees as well as medicinal species, each new complete mitogenome adds a data point that sharpens both phylogeny and our understanding of how fungal genomes evolve. The P. tuberculosus genome, published open access and citable with a permanent DOI, now joins that growing library, and the strongly supported Phellinus clade it anchors suggests that more genus-level mitochondrial comparisons across Hymenochaetaceae will be well worth watching.

Subject of Research: Comparative mitochondrial genomics of the medicinal fungus Phellinus tuberculosus within the family Hymenochaetaceae

Article Title: Comparative mitochondrial genomics of the medicinal mushroom Phellinus tuberculosus reveals conserved core genes and dynamic genome evolution within Hymenochaetaceae

Article References: Jin, C., Yuan, M.-X., Ma, J.-X., Ji, Y.-Y., Tang, L.-X., Li, X., Chen, M.-Y., Si, J., & Cui, B.-K. (2026). Comparative mitochondrial genomics of the medicinal mushroom Phellinus tuberculosus reveals conserved core genes and dynamic genome evolution within Hymenochaetaceae. BMC Genomics. https://doi.org/10.1186/s12864-026-13454-8

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13454-8

Keywords: Phellinus tuberculosus, mitochondrial genome, Hymenochaetaceae, comparative genomics, RNA editing, white-rot fungi, medicinal mushrooms, phylogenomics, simple sequence repeats, fungal evolution, BMC Genomics, genome annotation

The central question the researchers posed is deceptively simple: do enhancers share recurring, sequence-based features that distinguish them from the rest of the genome? The team refers to this hypothetical signature as enhancerness. If such a signature exists and can be learned by a machine, it would mean that the grammar of gene regulation is written in the sequence itself, in patterns of bases that recur across species, cell types, and experimental assays. Confirming that would be biologically fundamental, because it would suggest a universal logic underlying how regulatory information is encoded. It would also be enormously practical, since a model that recognizes enhancerness could annotate newly sequenced genomes without waiting for years of costly experiments. The study provides evidence on both fronts, combining computational benchmarks with validation in living flies.

EnhancerDetector is built on a convolutional neural network, the same family of architectures that revolutionized image recognition. Convolutional networks excel at detecting local patterns and assembling them into higher-order structures, which makes them well suited to DNA, where transcription factor binding sites and their combinations form the functional vocabulary of a regulatory element. Trained on human data, the model takes short sequence windows as input and directly outputs a probability that the window is an enhancer. This design choice matters. Many existing enhancer predictors rely on chromatin features, such as histone modifications like H3K4me1 and H3K27ac, which are measured in specific cell types and require complicated post hoc thresholding to convert into enhancer calls. By working directly from sequence, EnhancerDetector sidesteps that dependency and simplifies the discovery workflow, producing scores that can be ranked and filtered without additional processing layers.

The performance results are striking in their breadth. On human, mouse, and fly datasets, EnhancerDetector consistently outperformed existing methods in precision and F1 score, two metrics that matter greatly in genomics because the number of true non-enhancers vastly exceeds the number of true enhancers, making false positives the dominant error. The model also generalized across datasets generated with diverse experimental assays, suggesting it had not merely memorized the artifacts of one laboratory technique. An ensemble strategy, in which multiple models vote together, further improved reliability by reducing false positives. In a field where a single mislabeled enhancer can send a research group down an expensive dead end, that reduction in noise is a meaningful advance. The authors also compared their approach against sequence-based competitors such as gapped k-mer support vector machines, and the convolutional network came out ahead.

Perhaps the most practically important result concerns transfer learning. EnhancerDetector was trained on human data, yet it performed well on mouse and fly, species separated from humans by tens and hundreds of millions of years of evolution respectively. Even more remarkably, the model can be fine-tuned on a new species and retains strong performance when adapted with as few as 20,000 enhancer sequences. For newly sequenced genomes, where experimental data are scarce by definition, this is a game changer. A research consortium that has just assembled the genome of an obscure organism can adapt a pretrained model with a modest dataset and begin prioritizing candidate regulatory elements immediately, rather than starting from scratch. The authors position EnhancerDetector explicitly as a first-stage annotation tool for exactly this scenario.

Accuracy alone, however, was not the goal. The team wanted a model that biologists could interrogate, one whose predictions could be traced back to specific features of the DNA. To achieve this, they applied class activation maps, a visualization technique that highlights which regions of an input sequence contributed most strongly to the model’s decision. These maps allow researchers to see, base by base, where the network is focusing its attention when it declares a stretch of DNA to be an enhancer. Such interpretability transforms the model from a black box into a hypothesis generator: the highlighted motifs can be compared against known transcription factor binding sites, and unexplained patterns can point toward regulatory logic that science has not yet characterized. In an era when machine learning is often criticized for opacity, building interpretability into the architecture from the start is a deliberate and welcome choice.

Computational benchmarks are one thing, but the gold standard in genomics is experimental validation, and here the study delivers. The team tested candidate enhancers predicted by the model in transgenic flies, using reporter constructs in which a candidate DNA sequence is placed next to a green fluorescent protein gene, so that if the sequence acts as an enhancer, the tissue where it is active glows under the microscope. Of six candidates tested, five drove reporter expression, a hit rate that would be respectable even with extensive prior experimental support. Moreover, four of the five exhibited expression patterns consistent with those reported in the earlier literature, providing independent corroboration that the model was not merely finding sequences capable of driving expression in some generic way, but was recovering biologically meaningful regulatory elements with the correct tissue specificities.

Taken together, the computational and experimental results support the paper’s central claim: enhancers possess learnable, recurring sequence features that can be captured in one species and transferred to another. The analyses identified distinct sequence and contextual features associated with enhancer activity, the constellation the authors call enhancerness. This finding resonates with a long-standing debate in regulatory genomics. Enhancers are notoriously diffuse, with binding sites for multiple transcription factors arranged in flexible combinations rather than rigid templates, which has led some researchers to doubt whether any universal sequence signature could exist. The success of a single model trained on human sequence and applied to fly suggests that, beneath the flexibility, there is a shared statistical texture to enhancer DNA, one deep enough for a convolutional network to grasp but subtle enough to have eluded simpler analytical methods.

The implications extend well beyond the three species studied. As sequencing costs continue to fall, thousands of genomes are being assembled for organisms with no experimental annotation at all, from non-model animals to endangered species to agricultural pests. A tool that can flag candidate enhancers from sequence alone, and be fine-tuned with a few tens of thousands of examples, offers a scalable path to functional annotation of this flood of new data. The work was supported by the National Human Genome Research Institute of the National Institutes of Health under award number R21HG011507, and the article is open access, so the tool and its findings are available to any laboratory that wants to build on them. The authors acknowledge assistance with experimental validations from Jack Leatherbarrow and James Weiser, and they declare no competing interests.

There remain limits to what any sequence-only model can do. Enhancer activity depends on cellular context, and a sequence that functions in one tissue may be silent in another, so predictions from DNA alone will always need experimental follow-up. But by demonstrating that a common, learnable signature of enhancerness exists across species and assays, and by packaging that insight into an interpretable, transferable, and experimentally validated framework, the EnhancerDetector team has taken a substantial step toward making the dark matter of the genome a little less dark. For biologists staring at a freshly assembled genome and wondering where the switches are, the answer may now be one pretrained neural network away.

Subject of Research: Cross-species enhancer prediction from DNA sequence using interpretable convolutional neural networks

Article Title: EnhancerDetector: enhancer discovery from human to fly via interpretable deep learning

Article References: Solis, L. M., Sterling-Lentsch, G., Halfon, M. S., & Girgis, H. Z. (2026). EnhancerDetector: enhancer discovery from human to fly via interpretable deep learning. BMC Genomics. https://doi.org/10.1186/s12864-026-13262-0

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13262-0

Keywords: enhancers, deep learning, convolutional neural network, genome annotation, regulatory genomics, cross-species prediction, interpretability, class activation maps, Drosophila, transcription factor binding sites, transfer learning, BMC Genomics

Microproteins, often abbreviated MPs, are proteins encoded by small open reading frames, or sORFs, that typically produce peptides shorter than 100 amino acids. For decades these tiny coding sequences were dismissed as genomic noise, too short to be considered real genes and too small to yield easily detectable protein products. That view has changed dramatically over the past several years, as researchers working on humans, plants and model organisms discovered that microproteins act as regulators of processes ranging from muscle development to stress responses. Well-known examples include the LITTLE ZIPPER proteins, abbreviated ZPR, which modulate plant transcription factor networks. Yet in protozoan parasites, the functional roles and evolutionary conservation of microproteins have remained poorly understood, a gap the Taiwanese team set out to close for one of medicine’s most persistent pathogens.

Trichomonas vaginalis is a particularly compelling subject for this kind of survey. The parasite infects an estimated hundreds of millions of people worldwide, causes trichomoniasis, and is associated with increased risks of adverse pregnancy outcomes and heightened susceptibility to HIV infection. Its genome is famously large and repetitive, which complicates standard annotation pipelines and makes it a plausible refuge for uncharted small genes. Because microproteins are short and often weakly expressed, standard omics approaches struggle to distinguish genuine microprotein-coding genes from random stretches of sequence. The researchers therefore designed a multi-layered validation strategy specifically intended to overcome the limitations of conventional methods, layering independent lines of evidence so that each candidate microprotein had to survive several rounds of scrutiny.

The foundation of the analysis was transcriptomic. The team integrated bulk RNA sequencing data from 22 isolates of Trichomonas vaginalis, allowing them to ask which small open reading frames were consistently transcribed across genetically distinct parasite strains rather than appearing in only one background. This cross-isolate approach delineated what the authors describe as a stable core microproteome of 918 microprotein candidates expressed consistently across the isolates. Within that set, 786 genes correspond to core-expressed microprotein genes, while 132 candidates derive from non-annotated small open reading frames located in regions of the reference genome that had never been assigned a gene. The consistent length distribution of the expressed microproteins across all 22 isolates, ranging from very short peptides up to the 100 amino acid cutoff, underscores how stable the composition of this expressed microproteome is across the species.

One of the most striking findings concerns evolutionary novelty. When the team evaluated sequence homology between the parasite’s microproteins and the human proteome using stringent BLASTp criteria, they found that approximately 93 percent of the 918 candidates lack significant similarity to any human protein. That figure matters for two reasons. Scientifically, it suggests that the majority of these microproteins are either lineage-specific innovations or have diverged so far from their metazoan counterparts that sequence-based detection fails, making them a rich hunting ground for novel biology. Translationally, the scarcity of human homologs raises the possibility that many of these molecules could be targeted therapeutically with a reduced risk of cross-reacting with the human host, an important consideration for any future drug or vaccine development program built on this resource.

Functional predictions add further texture to the picture. About 14 percent of the identified microproteins were predicted to be membrane-associated or secreted proteins, a category of particular interest because membrane and secreted molecules sit at the interface between the parasite and its host and are prime candidates for involvement in host-parasite interactions, immune evasion and adhesion. Functional enrichment analysis of the broader set indicated that the microproteins are primarily associated with intracellular anatomical structures, ribosomal components and organelles, hinting at roles in fundamental cellular processes rather than peripheral functions. InterProScan annotation of the full candidate set, provided in the study’s supplementary tables, catalogs the domains, gene ontology annotations and pathway assignments that will let other researchers prioritize candidates for experimental follow-up.

Transcription alone does not prove translation, so the team turned to liquid chromatography coupled with tandem mass spectrometry, LC-MS/MS, to confirm that microproteins are actually made. Working with protein extracts from the isolates ATCC 30236 and PRA98, separated by SDS-PAGE and processed through both in-gel and in-solution digestion workflows, the researchers confirmed the translation of 103 microproteins. Each of these detections rests on peptide spectrum matches linking observed mass spectra to predicted microprotein sequences, providing direct physical evidence that the genome’s small open reading frames are not merely transcribed but actively translated into stable protein products. The validation results, including the breakdown of in-gel and in-solution identifications, are published as supplementary data accompanying the paper.

The third pillar of the validation strategy was single-cell RNA sequencing, employed to assess how robustly microprotein transcripts are expressed at the level of individual cells. Bulk RNA sequencing can mask heterogeneity within a population, so a transcript that appears abundant on average might in fact be produced by only a small subset of cells. The scRNA-seq analysis of isolate ATCC 30236 demonstrated that 149 of the microprotein candidates are detected in more than 50 percent of individual cells, indicating that their expression is a widespread property of the parasite population rather than a niche feature of a few cells. Single-cell prevalence data for the full candidate set are provided in the study’s supplementary tables, giving future investigators a quantitative measure of expression robustness to weigh alongside transcript abundance measured in transcripts per million.

The study’s authors emphasize that this work provides the first comprehensive atlas of microproteins in Trichomonas vaginalis and establishes a valuable resource for future functional investigations. That framing is worth taking seriously, because an atlas of this kind changes the economics of follow-up research. Instead of each laboratory having to rediscover which small open reading frames are real, expressed and translated, investigators can now start from a curated list of 918 candidates, of which 103 have mass spectrometry evidence and 149 have single-cell expression support, and select targets based on predicted localization, homology status and expression pattern. The supplementary tables, which include genomic locations, molecular weights, isoelectric points, start codons and nucleotide sequences, are designed to make that triage straightforward.

Looking ahead, the open questions are considerable. What do these microproteins actually do inside the parasite’s cells? Are the membrane-associated and secreted members of the set involved in the host-parasite interplay that determines infection severity? Do the 132 candidates from previously unannotated intergenic regions represent a class of genes that standard pipelines have systematically missed, not just in this parasite but in related protists? The Taiwanese team, supported by grants from the National Science and Technology Council and Chang Gung Memorial Hospital, has provided the map; the territory now awaits exploration. For a pathogen that has long been managed with a narrow arsenal of therapies and continues to pose public health challenges worldwide, the discovery that nearly a thousand tiny, mostly human-dissimilar proteins are woven into its biology is an invitation to look again at what a parasite genome really contains.

Subject of Research: Proteogenomic identification of microproteins encoded by small open reading frames in the protozoan parasite Trichomonas vaginalis

Article Title: A comprehensive proteogenomic landscape of microproteins in Trichomonas vaginalis

Article References: Syu, J.-W., Luo, H.-W., Lin, R., Huang, P.-J., Lee, C.-C., Yeh, Y.-M., Chiu, C.-H., Ong, S.-C., & Tang, P. (2026). A comprehensive proteogenomic landscape of microproteins in Trichomonas vaginalis. Parasites & Vectors. https://doi.org/10.1186/s13071-026-07669-7

Image Credits: AI Generated

DOI: 10.1186/s13071-026-07669-7

Keywords: Trichomonas vaginalis, microproteins, small open reading frames, proteogenomics, LC-MS/MS, single-cell RNA sequencing, transcriptomics, parasite biology, host-parasite interactions, genome annotation, sexually transmitted infection, Parasites & Vectors

The research team, led by Imad Abugessaisa of the RIKEN Center for Integrative Medical Sciences together with colleagues including Wendy van de Sande of Erasmus MC, built their atlas using a technique called LQ-ssCAGE, or low-quantity single-strand cap analysis of gene expression. CAGE-based methods capture the 5-prime capped ends of RNA molecules, which mark the exact positions where RNA polymerase II initiated transcription. Because the method both identifies these transcription start sites and quantifies how abundantly each one is used, it provides a two-in-one view of promoter activity and gene expression. The researchers prepared 72 LQ-ssCAGE libraries and, in parallel, 72 standard RNA-seq libraries from the same samples, using the RNA-seq data as an independent quality check on the CAGE measurements.

The experimental design was carefully constructed to mimic a real infection and its treatment. Half of the larvae were infected with Madurella mycetomatis, a high-priority fungal pathogen on the World Health Organization’s fungal priority list and the most common cause of eumycetoma, a neglected tropical disease characterized by tumorous lesions in subcutaneous tissue. The infected and uninfected larvae were then divided into three groups: one treated with solvent only, one with the antifungal drug itraconazole, and one with ravuconazole, which was recently assessed in the first clinical trial for eumycetoma. Samples were collected at four time points, 6, 30, 72, and 168 hours, with three biological replicates per condition, producing a rich time-course dataset of promoter activity during infection and therapy.

Sequencing yielded roughly 53.3 million cap-trapped 5-prime reads, which the team mapped to the current reference genome assembly, CSIRO_AGI_GalMel_v1, a high-quality assembly scoring 99.3 percent on the lepidoptera BUSCO benchmark and containing 15,138 genes and pseudogenes. Quality control was thorough. Gene body coverage showed the expected strong 5-prime enrichment for CAGE libraries and uniform coverage for RNA-seq libraries, sequencing depth was consistent across samples with no outlier batches, and pairwise correlation analysis between normalized RNA-seq and CAGE counts showed strong positive Spearman correlations, confirming that the two independent assays produced reproducible expression profiles.

From the raw data, the researchers identified 1,219,124 individual CAGE transcription start sites, which they filtered down to 332,601 biologically relevant positions. Clustering these signals produced 40,200 tag clusters, from which they derived two tiers of transcription start site peaks: a permissive set of 27,942 peaks detected in at least four samples, and a robust set of 5,437 peaks detected in all samples. They also identified 728 bidirectional clusters, the molecular signature of active enhancers, which after filtering to non-exonic intronic and intergenic locations yielded 141 candidate active enhancers. Users of the atlas can therefore choose between the permissive and robust annotations depending on how conservative their analyses need to be.

The validation evidence for these calls is strong. Among all CAGE peaks, 32.1 percent overlapped annotated 5-prime untranslated regions, and more strikingly, 80.27 percent of all CAGE tags fell within 5-prime UTRs, exactly where active promoters should sit. Genome-wide enrichment analysis showed that the identified peaks were strongly and specifically enriched at annotated promoters of both protein-coding and non-coding RNA genes, while intronic and intergenic regions were depleted. Dinucleotide frequency analysis at the dominant start sites revealed the characteristic initiator sequence bias expected at insect RNA polymerase II promoters, and the atlas-derived annotations displayed a markedly stronger core promoter signature, including TATA box and GC patch signals, than the existing NCBI reference annotation, demonstrating a genuine improvement in transcription start site resolution.

The atlas also revealed the two canonical shapes of animal promoters. Using the interquartile range of tag clustering as a measure of initiation precision, the team classified 4,746 broad promoters, where transcription starts at many scattered positions, and 1,679 sharp promoters, where initiation is concentrated at a single precise site. As expected from earlier CAGE studies in mammals and flies, sharp promoters carried a stronger TATA box positioned roughly 30 bases upstream of the start site, while broad promoters were GC-rich, consistent with CpG island-associated architecture. Most genes, about 6,899, used a single dominant start site, whereas roughly 2,300 genes used multiple sites, and 71.1 percent of promoter-associated genes were linked to more than one tag cluster, echoing the widespread alternative promoter usage documented in Drosophila.

Perhaps the most clinically intriguing result concerns what did and did not change during infection. RNA-seq confirmed differentially expressed genes between infected and healthy larvae, consistent with the team’s earlier work on iron regulatory pathways during M. mycetomatis grain development. However, no differentially expressed genes were detected between infected larvae and infected larvae treated with itraconazole or ravuconazole. The authors note that this mirrors recent clinical observations in which neither drug showed measurable activity during the first six months of eumycetoma treatment, with persistent fungal viability and no reduction in lesion size or plasma beta-1,3-D-glucan levels. Differential transcription start site usage analysis did flag a handful of genes, including a cystinosin homolog and a peroxidase-like isoform, whose promoter selection shifted between healthy and infected samples, hinting that infection may add a layer of transcriptional regulation beyond simple changes in gene abundance.

Using expression correlations between nearby elements, the team predicted physical interactions between enhancers and promoters, finding that 682 of 37,320 transcription start sites interacted with 156 of 218 active enhancers. Interacting start sites were expressed at significantly higher levels than non-interacting ones, and the analysis also uncovered super-enhancers, clusters of closely spaced bidirectional enhancers. De novo motif discovery on the top 5 percent of expressed start sites pointed to RFX family transcription factors, with the RFX1 ortholog showing 7.1-fold enrichment, a factor conserved across animals and known in flies as a key regulator of ciliary and sensory neuron genes. The authors are candid about limitations: only about 700,000 reads per sample were generated, and saturation analysis suggests the depth was sufficient for highly expressed promoters but likely underestimated weak enhancers. Even so, with all data deposited in NCBI GEO and Mendeley Data, and the annotations browsable in the ZENBU genome browser, this first draft start site atlas gives the growing G. mellonella research community a foundation for gene regulation studies comparable to the reference resources long available for human and mouse.

Subject of Research: Genome-wide mapping of transcription start sites in Galleria mellonella larvae during fungal infection and antifungal treatment

Article Title: Genome-wide mapping of the Galleria mellonella transcription start sites during fungal infection and treatment

Article References: Abugessaisa, I., Konings, M., Manabe, R.-I., Tagami, M., Severin, J., Hasegawa, A., Kawashima, T., Kinoshita, H., Noma, S., Takahashi, C., Verbon, A., Okazaki, Y., van de Sande, W. W., & Kasukawa, T. (2026). Genome-wide mapping of the Galleria mellonella transcription start sites during fungal infection and treatment. iScience, 29(10), Article 117675. https://doi.org/10.1016/j.isci.2026.117675

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117675

Keywords: Galleria mellonella, transcription start sites, CAGE, Madurella mycetomatis, eumycetoma, promoters, enhancers, gene regulation, itraconazole, ravuconazole, genome annotation, infection model

The four tools under scrutiny were Prokka, Bakta, EggNOG-mapper and PGAP, each representing a distinct philosophy of how to assign function to genes. Prokka, a long-standing workhorse of bacterial genomics, is prized for its speed and ease of use. Bakta, a more recent successor, extends the Prokka concept with expanded databases and richer functional annotation. EggNOG-mapper takes a different route, transferring functional information from precomputed orthology groups spanning thousands of organisms. PGAP, the NCBI’s Prokaryotic Genome Annotation Pipeline, is the pipeline behind the reference annotations deposited in public databases, giving it a special role in standardizing how prokaryotic genomes are described worldwide. Until now, no study had compared these tools head-to-head at anything approaching this scale.

The scale of the benchmark is what sets it apart. The team assembled a dataset of 156,033 genomes deliberately chosen to span the diversity and the messiness of real-world sequencing. At its core were Escherichia coli strains, which served as a baseline for performance because they are among the best-characterized organisms in biology, with densely curated gene catalogs against which predictions can be checked. Around that core, the researchers layered thousands of bacterial genomes from across the taxonomic spectrum, thousands of archaeal genomes representing a domain of life that is often underrepresented in benchmarking efforts, and, critically, categories of genomes that reflect the challenges of modern practice: artificially frameshifted genomes and metagenome-assembled genomes, known as MAGs, reconstructed from environmental sequencing data.

The inclusion of imperfect genomes is arguably the study’s most consequential design decision. In principle, annotation tools are developed and tuned on clean, complete, high-quality reference genomes. In practice, the genomes flowing through laboratories today are frequently fragmented assemblies, contaminated with foreign DNA, or stitched together computationally from short metagenomic reads, none of which ever saw a perfect reference. Frameshifts, in which small insertions or deletions disrupt the reading frame of a coding sequence, are a particular hazard, because a tool that fails to recognize a frameshifted gene may either miss the gene entirely or mislabel its function. By deliberately stress-testing the pipelines against these degraded inputs, the benchmark captures how the tools behave where it matters most: on the messy data that dominate real surveillance, clinical and environmental projects.

The headline result is that there is no single winner, and that the best choice depends on what kind of genome is on the bench. For high-quality bacterial genomes, Bakta emerged as the strongest performer, delivering the most reliable annotation when the input sequence was complete and well assembled. For archaeal genomes, however, the picture flipped: PGAP outperformed the alternatives, a finding with practical weight given that archaeal genomics has historically been underserved by tools optimized on bacteria. PGAP also proved the more robust option for challenging bacterial assemblies, including metagenome-assembled, fragmented and contaminated samples, suggesting that the NCBI pipeline’s underlying machinery handles imperfect input more gracefully than its competitors.

When the analysis turned to functional annotation, the assignment of Gene Ontology terms that describe what each gene product does, a further layer of nuance appeared. PGAP consistently provided broader term coverage, attaching Gene Ontology annotations to a wider share of features across the genome. EggNOG-mapper, by contrast, tended to assign more terms per annotated feature, painting a denser functional portrait of the genes it did annotate. These are not contradictory results but complementary ones, and they matter for anyone conducting enrichment analyses or building functional profiles from annotation output. A researcher whose analysis depends on maximizing the number of annotated genes may prefer broader coverage, while one seeking richer descriptions of individual proteins may favor the denser per-feature annotation that EggNOG-mapper provides.

The implications ripple outward through microbiology. Genome annotation is not a cosmetic step; it determines which genes are counted in resistance screens, which virulence factors are flagged in pathogen surveillance, which metabolic pathways are reconstructed from environmental MAGs, and which drug targets are proposed from newly sequenced isolates. If two laboratories annotate the same genome with different tools and reach different conclusions about its gene content, the reproducibility of the entire field is at stake. By quantifying exactly where the tools agree and diverge, and by tying those differences to genome quality, taxonomy and origin, the study gives researchers an evidence-based decision framework rather than a matter of habit or folklore. The authors also provide detailed supplementary data, including per-sample summaries for E. coli, archaeal, non-MAG bacterial and MAG bacterial groups, along with accession lists, so that other groups can apply the benchmark’s findings directly to their own data.

The study also carries a message for the developers of annotation software. Tools that excel on pristine reference genomes may quietly degrade on the fragmented and contaminated assemblies that increasingly dominate public archives, and a benchmark of this scale makes such weaknesses visible in a way that small-scale comparisons never could. The authors frame their work as informing future tool development, and the specific failure modes documented across frameshifted and metagenome-assembled genomes offer a concrete to-do list: better handling of disrupted reading frames, greater robustness to contamination, and improved performance on archaeal diversity are all areas where the benchmark identifies room for improvement. As sequencing costs continue to fall and MAG reconstruction becomes routine, the gap between the genomes tools were designed for and the genomes they actually receive will only widen unless development is guided by evidence of this kind.

Published open access in Genome Biology on 4 September 2026, the study arrives at a moment when the volume of prokaryotic sequencing data is growing explosively, driven by clinical surveillance, outbreak investigation, and large-scale environmental and host-associated microbiome projects. Every one of those projects begins with the same annotation step that this benchmark has now illuminated. For the working microbiologist, the practical guidance is straightforward: reach for Bakta when annotating high-quality bacterial genomes, trust PGAP for archaea and for assemblies that are fragmented, contaminated or metagenomic in origin, and weigh the trade-off between PGAP’s broader Gene Ontology coverage and EggNOG-mapper’s richer per-feature annotation when functional profiling is the goal. For the field as a whole, the study demonstrates that even the most established steps of the bioinformatics pipeline benefit from systematic, large-scale scrutiny, and it sets a template for how future annotation tools should be evaluated: not on a handful of model organisms, but across the full taxonomic and quality spectrum of the genomes scientists actually sequence.

Subject of Research: Large-scale benchmarking of prokaryotic genome annotation tools across diverse bacterial and archaeal genomes

Article Title: Large-scale benchmarking of prokaryotic annotation tools across thousands of species

Article References: Jundzill, M., Hölzer, M., Mangul, S., Marquet, M., Ehricht, R., Lohde, M., Spott, R., Makarewicz, O., Pletz, M. W., & Brandt, C. (2026). Large-scale benchmarking of prokaryotic annotation tools across thousands of species. Genome Biology, 27(1), Article 284. https://doi.org/10.1186/s13059-026-04262-0

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04262-0

Keywords: genome annotation, bioinformatics, bacteria, archaea, benchmarking, Prokka, Bakta, PGAP, EggNOG-mapper, metagenome-assembled genomes, Gene Ontology, computational biology

The problem that GLASS addresses arises after the initial alignment step. Long reads, particularly those produced by Oxford Nanopore sequencers, carry relatively high error rates compared with short reads. When these noisy reads are aligned to a genome, the aligner must decide whether apparent gaps represent genuine introns, consistent with known splice junctions, or simply sequencing artifacts. Misjudgments at this stage propagate downstream: transcript assemblers rely on splice junction signals within alignments to reconstruct full-length isoforms, and a single unreliable junction can corrupt an entire reconstructed transcript. The result is a catalog of predicted isoforms that mixes genuine biology with computational noise, undermining studies of gene regulation, disease mechanisms, and genome annotation.

GLASS, which stands for a graph learning algorithm for screening splice-aware alignments, takes aim at this problem with a post-alignment filtering strategy. Rather than modifying the aligners themselves, the method operates on their output, screening the alignments for reads whose splice patterns look suspicious. The central insight of the work is that the relationships between sequencing reads and splicing events form a rich network structure, and that this structure can be exploited by modern graph-based machine learning to distinguish trustworthy alignments from unreliable ones.

Technically, the method constructs what the authors call a Read–AS bipartite graph. In this graph, one set of nodes represents sequencing reads, and the other set represents alternative-splicing events, with edges connecting reads to the splicing events they support. This bipartite structure encodes a powerful form of collective evidence: a read that supports a splicing event also backed by many other reads gains credibility, while a read whose junctions appear in isolation, or in patterns inconsistent with the surrounding network, becomes suspect. Alternative splicing, the process by which a single gene can produce multiple RNA isoforms by joining exons in different combinations, generates exactly the kind of interconnected event landscape that such a graph can capture.

On top of this graph, GLASS deploys a BipartiteGCN model, a graph convolutional neural network adapted for bipartite structures. Graph convolutional networks work by allowing each node to aggregate information from its neighbors, layer by layer, so that a read node gradually accumulates a representation of its local splicing context. After several rounds of message passing, the network produces a classification for each read, flagging those with potentially unreliable or annotation-inconsistent splice patterns. This is a notable departure from conventional filtering approaches, which typically rely on simple per-read heuristics such as alignment scores or the number of supporting bases around a junction. By learning from the graph-wide context, GLASS can detect subtle inconsistencies that local rules would miss.

The authors evaluated GLASS across multiple datasets and several transcript assemblers, the software tools that convert aligned reads into reconstructed transcript models. The results showed a consistent pattern: applying GLASS reduced the burden of annotation-inconsistent splice signals in the alignments and improved the precision of conservative transcript reconstruction. In practical terms, the transcripts that survived GLASS filtering were more likely to be genuine, at the cost of a modest reduction in sensitivity, meaning some true but harder-to-confirm isoforms were also filtered out. This precision-oriented trade-off is a deliberate design choice, reflecting the view that for many applications, a shorter but cleaner catalog of isoforms is more valuable than a longer but noisier one.

Robustness checks formed an important part of the study. The team repeated their analyses with updated assemblers, with alternative genome annotations, and with alignments preprocessed by complementary tools including minisplice and 2passtools, which are designed to improve splice junction detection in long-read data. They also tested the approach on PacBio datasets, confirming that the benefits were not specific to one sequencing platform or preprocessing pipeline. Across these varied conditions, the same trend held: GLASS filtering pushed reconstruction results toward higher precision, supporting the claim that the method captures a general property of alignment quality rather than an artifact of any particular toolchain.

One of the most scientifically careful aspects of the paper is the authors’ explicit clarification of what annotation-inconsistent splice patterns actually mean. A splice junction that does not appear in a reference annotation is not necessarily an error. Long-read sequencing frequently uncovers genuine but previously unannotated isoforms, and some of the inconsistent signals flagged by GLASS may represent real biology that current annotation databases simply do not record. The authors therefore position GLASS not as an oracle that separates truth from falsehood, but as a complementary filtering strategy aimed specifically at conservative transcript reconstruction, where the goal is to build a high-confidence catalog of isoforms rather than to exhaustively discover every possible splice event.

This framing has practical consequences for the field. Genome annotation projects, clinical transcriptomics, and studies of disease-associated isoforms all depend on transcript reconstructions that researchers can trust. False splice junctions embedded in reconstructed transcripts can lead to phantom proteins being predicted, misestimated gene structures, and wasted experimental effort chasing artifacts. A filtering layer like GLASS, which slots in after alignment and before assembly, offers a way to raise the quality floor of these pipelines without requiring users to abandon their preferred aligners or assemblers. Because it is precision-oriented, it is best suited to applications where confidence in each reported isoform matters more than completeness.

The work also illustrates a broader trend in computational genomics: the migration of graph neural networks from social network analysis and recommendation systems into the life sciences. Biological data is inherently relational, whether it involves protein interactions, regulatory networks, or, as here, the web of evidence connecting sequencing reads to splicing events. The Read–AS bipartite graph at the heart of GLASS is a clean example of how a domain-specific structure can be translated into a representation that modern learning algorithms handle naturally. The BipartiteGCN model learns which patterns of shared support and shared inconsistency tend to characterize reliable alignments, effectively distilling the collective judgment of the entire dataset into a per-read quality assessment.

Funded by China’s National Key R&D Program and the National Natural Science Foundation of China, the study comes from a mathematics-oriented research group, underscoring how interdisciplinary the field of genomics has become. As long-read sequencing continues to drop in cost and expand into clinical and population-scale projects, the volume of RNA-seq data demanding careful splice-aware processing will only grow. Methods like GLASS suggest a future in which the noisy, error-prone raw signal of long reads can be progressively refined by layers of intelligent filtering, yielding transcript catalogs accurate enough to support everything from basic gene discovery to the interpretation of disease variants. For now, the authors’ measured conclusion stands: GLASS is a complementary tool, one more safeguard in the long chain of computation that turns raw molecular signals into biological knowledge.

Subject of Research: Graph-learning-based filtering of splice-aware alignments in long-read RNA-seq transcriptomics

Article Title: GLASS: a graph learning algorithm for screening splice-aware alignments of long-read RNA-seq data

Article References: Li, J., Tan, Z., Li, G., & Yu, T. (2026). GLASS: a graph learning algorithm for screening splice-aware alignments of long-read RNA-seq data. BMC Genomics. https://doi.org/10.1186/s12864-026-13334-1

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13334-1

Keywords: long-read RNA-seq, alternative splicing, graph neural network, transcript reconstruction, splice-aware alignment, BipartiteGCN, read classification, genome annotation, Oxford Nanopore, PacBio, bioinformatics, machine learning

The stakes of such an investigation are considerable. Tetracycline antibiotics work by binding to the 30S ribosomal subunit of bacteria, blocking the attachment of aminoacyl transfer RNA and thereby halting protein synthesis. Decades of clinical and agricultural use have driven the spread of resistance across bacterial communities, and lactic acid bacteria used in food fermentations are no exception. Because probiotic strains are consumed live and may reside temporarily in the gut, any resistance genes they carry could, in principle, be transferred to commensal or pathogenic bacteria. The European Food Safety Authority therefore sets microbiological cut-off values for antibiotic susceptibility in microorganisms intended for food or feed applications, and strains exceeding those thresholds face heightened scrutiny before they can be considered safe.

That is precisely the situation with L. buchneri KU200793, a strain previously isolated from Korean fermented foods and characterized for its probiotic properties and neuroprotective effects. When the research team, led by corresponding author Young-Seo Park, measured the minimum inhibitory concentration of tetracycline against the strain using standardized methods for non-enterococcal lactic acid bacteria, they arrived at a value of 128 milligrams per liter. That figure exceeds the EFSA cut-off, meaning the strain would formally be classified as resistant and would need a mechanistic explanation before any safety assessment could proceed. The question the researchers posed was deceptively simple: is this resistance intrinsic and benign, or does it involve genes that could spread?

To answer it, the team first turned to the genome. Annotation of the KU200793 sequence revealed a suite of genes plausibly linked to antibiotic tolerance, including transporters belonging to the major facilitator superfamily and to the ATP-binding cassette family, a PmrA-type protein, and a transcriptional regulator of the TetR family. These categories are familiar players in resistance biology. Major facilitator superfamily pumps and ABC transporters can expel antibiotics from the cell before the drugs reach their ribosomal targets, while TetR-family regulators commonly control the expression of efflux systems in response to tetracycline itself. Notably, the analysis did not identify the classic acquired tetracycline resistance determinants, such as ribosomal protection genes of the tet(M) type or enzymatic inactivation genes, which are the elements most often mobilized horizontally between bacterial species.

The genomic picture was reinforced by comparisons at the drug’s actual target. Tetracycline resistance in many bacteria arises through mutations in the 30S ribosomal protein S10 or in the 16S ribosomal RNA that alter the antibiotic binding site. When the researchers compared these sequences in KU200793 with those of the tetracycline-susceptible reference strain L. buchneri ATCC 4005, they found no mutations at all. The ribosome of the resistant strain, in other words, appears structurally unremarkable, which pointed the investigation away from target modification and toward processes acting elsewhere in the cell.

One alternative hypothesis involved chemical inactivation of the drug itself. Some bacteria degrade or modify antibiotics extracellularly, rendering them harmless before uptake. The team used high-performance liquid chromatography to examine whether components released by tetracycline-exposed cultures of KU200793 affected the stability of the drug in the surrounding medium. The results suggested that extracellular components from exposed cultures may influence tetracycline stability, but the evidence did not support direct enzymatic inactivation as the mechanism. Whatever was happening outside the cell, it was not the straightforward destruction of the antibiotic molecule that characterizes classic inactivation-based resistance.

The decisive experiment came from pharmacological dissection of efflux. Efflux pumps are powered by the proton motive force or by ATP hydrolysis, and certain compounds can collapse that energy supply and disable the pumps. Chlorpromazine, a compound long used as an inhibitor of efflux-mediated resistance in studies dating back to work on fluoroquinolone-resistant Staphylococcus aureus, was applied to KU200793 alongside tetracycline. The effect was dramatic: the minimum inhibitory concentration of tetracycline dropped sixteen-fold, from 128 milligrams per liter to just 8 milligrams per liter. A sixteen-fold reduction upon efflux inhibition is strong functional evidence that active export is the dominant determinant of the resistance phenotype, converting a strain that exceeds regulatory thresholds into one that would fall within them if its pumps were silenced.

Yet the story did not end with efflux. The researchers also probed cell envelope permeability by measuring alkaline phosphatase activity, an enzyme whose accessibility to substrates depends on the outer layers of the Gram-positive cell wall. Reduced alkaline phosphatase activity in the resistant strain provided supportive evidence that decreased permeability contributes to the phenotype, limiting how much tetracycline enters the cell in the first place. Resistance built on restricted permeability is considered less worrisome from a horizontal gene transfer standpoint than resistance encoded by discrete, mobile resistance genes, because it reflects the strain’s own architectural and physiological traits rather than an acquired genetic element.

The final layer of evidence came from an elegant cell-free translation assay. Using a system in which green fluorescent protein is synthesized outside a living cell, the team tested whether cellular factors from KU200793 could restore translation in the presence of tetracycline. Partial restoration of cell-free GFP synthesis indicated that inducible translation-associated factors contribute to the strain’s ability to keep making proteins under antibiotic pressure. Together with the efflux and permeability findings, this positions the resistance of KU200793 as a multifactorial phenotype in which no single mechanism is solely responsible, but in which chlorpromazine-sensitive efflux clearly plays the leading role.

For the probiotics industry, the implications are twofold. On one hand, the absence of canonical acquired resistance genes, ribosomal target mutations, and direct drug inactivation is reassuring: the strain’s tolerance appears rooted in its intrinsic physiology rather than in a transferable genetic package. On the other hand, the study demonstrates the depth of characterization that modern safety assessment demands, and it offers a template for how genome annotation, inhibitor-based functional tests, and biochemical assays can be combined to distinguish benign intrinsic tolerance from genuine risk. As candidate probiotic strains continue to move from fermented foods into regulated applications, mechanistic studies of this kind will increasingly determine which organisms earn a place in the food supply, and the case of L. buchneri KU200793 shows that a sixteen-fold drop in resistance with a single inhibitor can speak louder than any genome annotation alone.

Subject of Research: Tetracycline resistance mechanisms in the probiotic bacterium Lentilactobacillus buchneri KU200793

Article Title: Molecular characterization of tetracycline resistance mechanisms in the probiotic strain Lentilactobacillus buchneri KU200793

Article References: Kang, Y., Jeong, H., Kang, D.-K., Paik, H. D., & Park, Y.-S. (2026). Molecular characterization of tetracycline resistance mechanisms in the probiotic strain Lentilactobacillus buchneri KU200793. Food Science and Biotechnology. https://doi.org/10.1007/s10068-026-02296-5

Image Credits: AI Generated

DOI: 10.1007/s10068-026-02296-5

Keywords: probiotics, Lentilactobacillus buchneri, tetracycline resistance, antibiotic resistance, efflux pumps, lactic acid bacteria, food safety, EFSA, minimum inhibitory concentration, chlorpromazine, genome annotation, cell envelope permeability