Uchimata is a toolkit for visualization of 3D structures of genomes. It consists of two packages: a Javascript library facilitating the rendering of 3D models of genomes, and a Python widget for visualization in Jupyter Notebooks. Main features include an expressive way to specify visual encodings, and filtering of 3D genome structures based on genomic semantics and spatial aspects. Uchimata is designed to be highly integratable with biological tooling available in Python.

Venoms from ant (Formicidae) are chemically extraordinarily diverse, yet their genomic architecture and evolutionary dynamics remain opaque. The "aculeatoxin hypothesis" assumes that most venom peptides in stinging insects (aculeate) evolutionarily derive from a single gene family, the aculeatoxins. The recent refutation of this hypothesis for bees raises fundamental questions about venom evolution in ants, which contribute most aculeatoxins. To trace venom peptide evolution, we performed synteny-aware comparative genomics across 25 ant species spanning major subfamilies. We analyzed phylogeny through sequences, structures, and embeddings from protein Language Model (pLMs).  We identified three conserved genomic regions (GR1-GR3) as evolutionary hotspots for ant venom genes, each exhibiting distinct evolutionary dynamics. Most remarkably, we discovered genuine melittin orthologs in ants at the conserved bee syntenic position (GR2), pushing the origin of this scaffold back to early aculeates, with a possible origin deeper in Hymenoptera. Gene copy numbers vary dramatically (0-17 genes per region), with predatory species showing expansions and formicine ants (subfamily Formicinae, which rely on formic acid spraying) showing reductions. Twenty-two distinct toxin clades emerge, with region-specific distributions suggesting repeated recruitment to conserved platforms.

Ants evolutionarily succeed by combining single-copy conservation (bee-like at GR2), massive gene duplication (snake-like at GR1), and repeated lineage-specific recruitment to conserved genomic platforms (GR3). This multi-modal evolution on stable genomic scaffolds, evidenced here, reconciles previously conflicting models of venom evolution and reveals how genomic architecture constrains, yet repeatedly enables, molecular innovation. Our findings highlight a general principle of genome evolution: complex adaptive traits can arise not from a single origin or mechanism, but through recurrent reuse of permissive genomic loci shaped by ecology. Such principles are likely relevant not only to other venomous animals, but more broadly to the evolution of complex multi-genic traits.

Optimizing synonymous codon sequences to improve translation efficiency, RNA stability, and compositional properties is challenging because the search space grows exponentially with protein length and objectives interact through long range RNA structure. Dynamic programming-based methods can provide strong solutions for fixed objective combinations but are difficult to extend to additional constraints. Deep generative models require large-scale, high-quality mRNA sequence datasets for training, limiting applicability when such data are scarce. Reinforcement learning naturally handles sequential decision-making but faces challenges in codon optimization due to delayed rewards, large action spaces, and expensive structural evaluation. We present CodonRL, a reinforcement learning framework that learns a structural prior for mRNA design from efficient folding feedback and demonstration-guided replay, and then enables user-controlled multi-objective trade-offs during inference. CodonRL uses LinearFold for fast intermediate reward computation during training and ViennaRNA for final evaluation, warms up learning with expert sequences to accelerate convergence for global structure objectives, and introduces milestone-based intermediate rewards to address delayed feedback in long range optimization. On a benchmark of 55 human proteins, CodonRL outperforms GEMORNA, a state-of-the-art codon optimization method, across multiple metrics, achieving 9.5% higher codon adaptation index (CAI), 25.4 kcal/mol more favorable minimum free energy (MFE), and 3.4% lower uridine content on average, while improving codon stabilization coefficient (CSC) in over 90% of benchmark proteins under matched constraints. These gains translate into designs that are predicted to be more efficiently translated, more structurally stable, and less immunogenic, while supporting continuous objective reweighting at inference time.

Flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) are essential redox cofactors in cellular metabolism. In this study, riboflavin-producing E. coli was engineered for de novo FMN and FAD synthesis. The RibFM-1 mutant with defective FAD synthase activity was overexpressed to produce FMN. NADH supply was enhanced by overexpressing sthA and deleting pntAB, while deletion of ndh encoding NADH dehydrogenase II and appB encoding cytochrome oxidase bd-II improved ATP regeneration. The resulting strain FMN08 produced 3457.3 mg/L FMN in fed-batch fermentation. FMNAT from Candida glabrata was engineered by semirational design to increase catalytic activity 2.14-fold, and coexpression of FMN1 from Saccharomyces cerevisiae enabled FMN-to-FAD conversion. Deletion of ushA encoding 5′-nucleotidase and overexpression of purA further promoted FAD production. The final strain FAD13 produced 2305.2 mg/L FAD in fed-batch fermentation. These results represent the highest reported titers of de novo synthesis of these redox cofactors.

It has long been debated in ecology whether communities behave as cohesive units or as loose collections of independent species. Here, we study this question in the context of community coalescence, the mixing of previously isolated communities, using bacterial microcosm experiments combined with ecological modeling. Our results demonstrate that interspecies interaction strength determines whether communities or species are the units of selection during coalescence. When interactions are moderate to strong, one parental community consistently outcompetes the other, indicating community-level selection. In contrast, under weak interactions, species fates are uncorrelated and the two communities contribute equally to the coalesced outcome, indicating the absence of community-level selection. These patterns extend to communities derived from natural samples with greater taxonomic diversity and richness. Furthermore, we identify two distinct regimes underlying community-level selection in experiments with different media conditions: an emergent regime in which collective dynamics shape outcomes that cannot be predicted from species traits alone, and a top-down regime where dominant species determine the winning community. Together, these results reconcile conflicting observations on community-level selection during community coalescence by demonstrating that communities behave as cohesive units only when interactions are sufficiently strong.

Root exudates play a central role in nutrient cycling, microbial recruitment, and plant-plant interactions, yet most experimental approaches for analyzing exudate chemistry rely on sterile hydroponic systems that poorly represent soil conditions. We present a low-cost, open-source, 3D-printed rhizobox platform and associated workflow that enable non-destructive root imaging and targeted rhizosphere soil sampling for LC-MS based metabolomics under realistic soil conditions. The design integrates a transparent removable window for repeated root observations, a defined soil volume to support spatially explicit sampling, and a blank-informed data-processing pipeline to distinguish plant-derived metabolites from soil and construction material background. We validated the system using the model plants Arabidopsis thaliana (Col-0) and Phragmites australis. We demonstrate reliable plant growth and consistent root development across the imaging window. We also show robust detection of species-specific rhizosphere metabolite profiles, with minimal variation in the vertical or temporal dimensions relative to the strong species effects. We further illustrate the application of the workflow in a factorial experiment manipulating social context (solo vs. conspecific pairs) and short-term heat stress in A. thaliana, showing that the approach is sensitive to treatment-associated changes in metabolite richness, diversity, and chemical composition in soil. The complete protocol, from rhizobox fabrication and assembly to soil extraction, LC-MS acquisition, and data curation can be implemented within 4-6 weeks using standard laboratory equipment and openly available design files. By combining ecological realism with analytical control, this workflow provides a broadly applicable method for quantifying rhizosphere metabolite dynamics across species, treatments, and spatial sampling zones, facilitating experimental studies of below-ground chemical processes in plant ecology.

Predicting enzymatic function from a protein sequence is a fundamental task in protein discovery and engineering. In this paper, we present Semi-supervised Learning for Enzyme Classification (SLEEC): a semi-supervised learning framework that learns a function-aware protein representation for Enzyme Commision (EC) number prediction. SLEEC achieves SOTA performance on standard benchmarks and provides interpretable, residue-level annotations. We further demonstrate that our framework is robust to benign sequence modifications routinely observed in protein engineering workflows-such as appending functional tags- a desirable property that current ML frameworks lack. Our main technical contribution is a multiple sequence alignment (MSA)-based data augmentation technique for discovering sparse residue activations within a given enzyme sequence.