Adam Paul Arkin

Agriculture is both humanity’s largest infrastructure and the leading driver of environmental degradation. Many early-stage biotechnologies could mitigate this impact, but their deployability has not been systematically evaluated. We analyze deployability of 77 biotechnologies across eight domains (crops, soils, livestock, watersheds, aquaculture, health and security, food loss/waste and frontiers) by applying a Balanced Readiness Level assessment spanning technological, market, regulatory, acceptance and organizational dimensions. We find that market viability and social acceptance, not technological readiness, represent the primary constraints. High-readiness technologies tend to be ‘drop-in’, contained, transient and perceived as natural. These characteristics indicate strategies to elevate overall deployability, such as addressing genetically modified organism containment. Assessing the scalability and environmental impacts of biotechnologies for prioritization requires additional quantitative analysis, as shown in a Fermi analysis of microbial fertilizers. We combine our findings into a near- to long-term roadmap for living infrastructure: predictively designed multicomponent biological communities that function as systems at the ecosystem level to enhance food security and environmental resilience.

Single-stranded RNA (ssRNA) and single-stranded DNA (ssDNA) bacteriophages form a key component of the global virome, yet the host genetic networks supporting their infection cycles remain poorly understood. Here we present a genome-wide map of the host factors involved in infection by F pilus-dependent ssRNA and ssDNA phages in Escherichia coli. Screens across ssRNA phages spanning all four genogroups of the Leviviricetes reveal a conserved network of host dependencies, with the notable exception of the F plasmid gene traD. Primary structural receptor components and DsbA-mediated disulfide bond formation are universally required to ensure F pilus integrity, whereas traD shows a strict genogroup-specific requirement during entry despite all lineages sharing the same receptor. Gene dosage screens identify elevated copy number of the genomic fragments encoding the HslU protease or the RNA chaperone StpA as barriers that restrict infection. Parallel assays with filamentous ssDNA phages confirm reliance on the host TolQRA complex for entry downstream of pilus engagement and track negative fitness signatures across envelope homeostasis clusters, reflecting the physiological burden of non-lytic virion extrusion. Comparative genomics and high-throughput genetics trace the selectivity of the isolation host, E. coli HSF, to a horizontally acquired Klebsiella capsule that excludes a large panel of double-stranded DNA phages isolated on other E. coli strains while allowing ssRNA and ssDNA virions to engage the extended F pilus and bypass the barrier through native pilus retraction. The work delivers a systematic, class-wide map of single-stranded phage-host interactions, a host platform for accessing uncultured viral diversity, and a functional blueprint for diagnostics, protein antibiotics and biocontrol tools that halt horizontal gene transfer.

The ModelSEED Biochemistry Database ( https://modelseed.org/biochem ) supplies foundational mass- and charge-balanced reaction networks for metabolic reconstructions. Here we present an update on the biochemistry where the database has been expanded to ∼ 46, 000 compounds, featuring ∼37, 000 metabolic structures, and ∼48, 000 reactions. We expanded our approach for handling thermodynamic data, enabling multiple sources of data to be derived, integrated, and presented to the wider research community. We now publish predictions of pKa, reaction energy (and respective uncertainties), and estimates of reaction direction from multiple independent sources. Reaction directions are graded gold, silver, or bronze according to the strength of supporting evidence, which is complemented by the prediction from an ensemble of large language models. These diverse sources empower users to understand agreements or discrepancies in directionality for ∼27, 000 reactions. Finally, to ensure data integrity, a new conflict-resolution pipeline reconciles structures across sources, documenting input from curators. Our work is publicly available at https://github.com/ModelSEED/ModelSEEDDatabase .

Microbial carbon utilisation shapes ecological niches, community assembly, and nutrient cycling, yet remains difficult to predict from genomes despite well-characterised pathways. Genome-based machine-learning models often generalise poorly across datasets, a limitation commonly attributed to small or taxonomically biased training sets. We integrated binary growth phenotypes for 819 strains across 240 carbon sources from four datasets, focusing on 15 shared sources. Cross-dataset evaluation reduced balanced accuracy from 0.86 (within datasets) to 0.62, and restricting evaluation to in-clade relatives recovered only 0.03 of this loss. Because phylogeny alone could not explain the transfer gap, we restricted training to samples with coherent genotype-phenotype relationships, defined by agreement between experimental labels and annotated pathways. Concordance filtering doubled the number of carbon sources with reproducible genomic signals across datasets (from 6 to 12 of 15), often recovering known pathway genes, while matched random subsets remained at the unfiltered level. The smaller concordant set also approached the cross-dataset performance of models augmented with up to 1800 BacDive genomes per phenotype, demonstrating greater efficiency per sample. Concordance-trained models remained bounded specialists, but separated growers from non-growers among genomes the pathway rules called negative, mostly generalists with broad-specificity transporters. Model confidence made these bounds actionable, since abstention improved accuracy where models transferred and retraining on newly labelled low-confidence genomes gave the largest gains for weaker phenotypes. Collectively, these findings recast generalisation as a problem of label-mechanism agreement and demonstrate how genome-based predictions can be deployed within explicit applicability domains and improved through targeted phenotyping.

For years, the airway microbiota have been theorized to be gatekeepers of respiratory health, as pathogens entering the airway make contact with resident microbes prior to or coincident with their interaction with host cells. Thus, modification of the native airway community may serve as a means of altering the local environment in favor of health. In this work, we hypothesize that synthetic bacterial communities introduced into the airway can serve as prophylactic countermeasures against infection by Burkholderia thailandensis in mice. We demonstrate that understanding of antagonistic interactions between a pathogen and airway microbiota in vitro can guide identification of probiotics with protective capabilities in vivo. Specifically, we show that niche overlap between the probiotic and pathogen is indicative of probiotic performance in vivo. This work serves as a foundation for the rational design of probiotic communities for protection against and treatment of respiratory infections.

The outermost surface layers of Gram-negative bacteria determine phage access to terminal receptors, yet their genetic basis has been mapped almost exclusively in laboratory strains that lack them. Here we apply genome-wide RB-TnSeq fitness profiling to four Escherichia coli O157:H7 strains from distinct phylogenetic clades sharing the O157 O-antigen, using 38 phages with terminal receptors previously mapped in E. coli K-12 strain. RB-TnSeq fitness landscapes across all four pathogenic backgrounds were mostly similar, and dominated by surface-associated loci, including the gfc-etk group 4 capsule operon, O-antigen biosynthesis genes, LPS core assembly genes and outer membrane proteins. Disruption of gfc-etk abolished infection in 11 genetically diverse myoviruses, establishing the O-antigen capsule as a widespread required primary recognition substrate. O-antigen loci generated two classes of fitness score patterns. For 10 phages, disruption increased infectivity, indicating it is a barrier to receptor access; for 3 others, disruption abolished infectivity, demonstrating it can also be a primary recognition substrate. Outer membrane protein receptor identity was conserved across laboratory and pathogenic backgrounds, with the same proteins recognized in both K-12 and O157:H7, while glycan layer state determines whether these receptors are reached. These results demonstrate that outer surface glycan layers can act as primary and optional recognition substrates for phage infection, or as physical barriers preventing terminal receptor access. Extending the ability to probe phage-targeted receptors beyond outer membrane proteins provides a framework for incorporating glycan layer state into predictive models of phage-host interactions.

The interrogation of data across biological and environmental systems has become increasingly complex. Fortunately, communities are adopting the FAIR (Findable, Accessible, Interoperable, Reusable) data principles for individual datasets, and continue to develop domain-specific, machine-actionable standards. However, integrating FAIR data for meta-analysis across data resources is still challenging. Understanding how disparate datasets are organized remains a manual, time-consuming process. Updating FAIR databases to reflect changes in knowledge is slow, allowing stale annotations and incorrect relationships to propagate, amplified by Artificial Intelligence (AI) systems that harvest data. Building on FAIR, we argue that data should be iteratively updated and improved. FAIR + COPE (Comparable, Organized, Predictive, Engaged) takes FAIR data and makes it Comparable, rapidly Organized (applying / updating standards) for Predictive models, which can be validated and improved by an Engaged community. We provide examples of FAIR + COPE resources and science use cases that highlight the importance of FAIR + COPE in scientific research.

How a single gram of soil harbors billions of microorganisms, each with distinct genomic variants that collectively maintain coherent ecological function(s), is one of microbiology's grand unsolved problems. A key obstacle is determining which variants contribute to individual- and community-level fitness, in which contexts, and how co-occurring ecotypes interact to divide niche space. Here, using nitrate (NO ₃ ^- )-contaminated subsurface sediment as inoculum, we have performed high throughput enrichments in laboratory media of defined carbon source compositions across ecologically relevant gradients of pH and NO ₃ ^- . Long-read metagenomics and link-community decomposition of co-occurrence networks of taxa across these enrichments has revealed context-specific functional interactions among dominant generalist and lower-abundance specialist denitrifier ecotypes that comprise 53 distinct enriched communities (EnComs) across 288 enrichments derived from a single sediment sample. We identified a single enzymatic difference of alternative NO ₃ ^- reductases (NapAB vs. NarGHI) with differing substrate affinities that provided a mechanistic explanation for competitive niche partitioning between the two dominant taxa, Neorhizobium spp. and Allorhizobium spp., along the NO ₃ ^- gradient. Genome-wide polymorphism ratios (pN/pS) revealed that selective pressures vary systematically with carbon source availability and gradients of pH and NO ₃ ^- , which helps explain the natural biodiversity and functional interactions of ecotypes within denitrifying communities in the subsurface sediment. Our findings show that controlled enrichments along ecological gradients can thus uncover eco-evolutionary forces of selection, drift, and diversification that sculpt the biodiversity of microbial populations in the natural environment.