Connecting mathematical models with empirically measured microbial growth has remained challenging, as numerous competing models based on different theoretical approaches can fit observations. Therefore, we develop a method to automatically propose growth models from microbial data alone. We validate this approach using an available dataset of E. coli grown on known resources, and study 14 species across various concentrations of a rich medium. The inherently interpretable approach of symbolic regression infers explicit dynamical models directly from growth data. Using symbolic regression natively, does not favor biologically interpretable models, but we find cumulative population gain to be a more informative machine learning feature than population size. Random Forest machine learning allows us to relate this finding to the approximation of a constant-rate per capita resource consumption. This suggests that the area under the growth curve (AUC) measured in routine experiments provides information on the effective resource dynamics governing microbial growth. Finally, we use theoretical insights to inform the symbolic regression algorithm and favor biologically interpretable models. Overall, we found that balancing between data fit, parsimony and biological relevance favored both the simplest, linear approximation and models based on Monod dynamics, with either one or two underlying resources. Therefore, our approach to read growth laws off of microbial batch cultures provides insights on data-driven modelling.

Limiting ribosome synthesis and activity is crucial for adaptation to stresses, such as heat or nutrient starvation. In Bacillus subtilis, this can be achieved through the coordinated action of the alarmones (p)ppGpp and the transcription factor Spx. Here, we performed a genetic screen to identify novel factors that contribute to the heat shock response in B. subtilis. We identified the Y-complex, which confers specificity to the endonuclease RNase Y, as a critical player under stress conditions, such as heat or transition into the stationary phase. This protein complex is required for the targeting and processing of diverse RNAs, notably the maturation of mRNAs encoding proteins involved in translation and metabolism. We further demonstrate that the Y-complex and RNase Y initiate the degradation of rRNAs of mature ribosomes, lowering their abundance. We propose that the Y-complex is a regulatory hub that modulates gene expression, adjusts protein synthesis and resource allocation. Transcriptional regulation is a central mechanism underlying bacterial adaptation to stress. Here, the authors used a genetic fitness screen to show that the Y-complex in Bacillus subtilis contributes to the heat shock response through its interaction with the ribosome.

N-linked glycosylation is critical for protein function and stability, yet identifying glycosylated sites remains challenging because glycosylation depends on sequence motifs and structural context. Many available computational approaches focus on motif-centred sequence windows and provide limited support for whole-protein inspection of candidate sites. SGGly is a freely accessible web server for structure-guided analysis of candidate N-linked glycosylation sites across full-length proteins. The server uses ProtBERT transformer-based embeddings with sequon and structure-derived residue descriptors to generate residue-level candidate-site predictions and returns downloadable residue-level predictions together with interactive 3D visualisation. Using a dual evaluation framework, SGGly achieved a Matthews correlation coefficient of 0.888 and receiver operating characteristic area under the curve of 0.987 under a strict, publication-supported regime. On the independent N-GlyDE benchmark, SGGly demonstrated strong generalisability, achieving the strongest specificity (0.941), sensitivity (0.993), and accuracy (0.946) among compared methods. SGGly provides a practical web resource for whole-protein glycosylation candidate mapping, structural inspection, and prioritisation of sites for follow-up analysis, guiding experimental design and interpreting glycoproteomic observations. 

Cas13-based RNA effectors may enable dynamic, multiplexed and reversible gene regulation in bacteria. Yet, their widespread adoption is hindered by inherent cytotoxicity and collateral cleavage. Here we present a rational protein engineering strategy to generate attenuated Cas13d variants with tunable RNase activity through targeted truncation of flexible regions. This permits effective transcript knockdown while greatly reducing toxicity as reflected by a 2.2-fold higher growth optical density. By introducing proximal mismatches at the 5′ end of CRISPR RNA spacers, our system allows functional switching between translation inhibition, polycistronic mRNA degradation and IF3-fusion-based translation-level CRISPR activation. We demonstrate programmable, orthogonal and multiplexed regulation of individual genes within polycistronic mRNAs and synthetic circuits. Application to lycopene biosynthesis optimization shows robust pathway rewiring and improved yields alongside fine-tuned modulation of essential and competing pathways in E. coli. Our work provides a versatile RNA-regulatory toolkit for next-generation microbial synthetic biology and RNA-based biotechnology. A Cas13-based gene switch allows multiplexed RNA regulation in bacteria.

Symbiotic bacteria in the gut have an important physiological impact on their hosts, but the mechanisms that underlie the exchange of molecular signals remain poorly understood. Membrane vesicles have been suggested to mediate the direct exchange of cytoplasmic content like nucleic acids (NAs), but their study is complicated by conflicting and imprecise reports of their type and composition. Here, we show that honeybee gut symbionts produce non-lytic membrane vesicles (MVs) enriched in NA, potentially explaining the RNAi activity of engineered Snodgrassella alvi in honeybees. Using cryogenic electron microscopy (cryo-EM), we developed a method to distinguish lytic from non-lytic MV production in Gram-negative bacteria and to differentiate outer membrane vesicles (OMVs) from outer-inner membrane vesicles (OIMVs) based on membrane ultrastructure. Among the strains studied, S. alvi, Gilliamella apicola, and Gilliamella apis exhibit clear OMV and OIMV budding, while Escherichia coli and Salmonella enterica show membrane debris and self-assembled vesicles, indicating lytic release. MVs from the symbionts carry significantly more NAs than non-symbionts. Assays on DNA and RNA contents confirm the cytoplasmic origin of MV cargo in S. alvi, suggesting a role in mediating NA delivery to the insect host. These findings enhance our understanding of symbiotic vesiculation and highlight the potential for engineering symbionts to boost honeybee immunity and deliver NA-based therapeutics via vesicular transport. Here, using cryo-EM, the authors visualize non-lytic outer and inner membrane vesicles budding from Gram-negative symbionts, and provide evidence that these nucleic acid-rich double membrane vesicles may transport and deliver cytoplasmic cargo to the host.

Nanopore adaptive sequencing enables real-time target enrichment, yet current deep-learning methods require costly, sample-specific experimental training data. To address this, we developed GANBase, a genome-guided generative adversarial learning framework, which is trained exclusively on reference sequences and incorporates a Monte Carlo Tree Search-based Rollout strategy for model training. GANBase demonstrates robust performance in target enrichment and host depletion across diverse scenarios. In live adaptive sequencing experiments, it remains effective despite significant pore loss or flow cell version updates, providing a data-independent solution that significantly expands the utility of real-time targeted sequencing. Adaptive nanopore sequencing enables targeted enrichment, but current methods rely on training data. Here, the authors develop GANBase, a genome‑guided deep‑learning framework that achieves robust, data‑independent target enrichment and host depletion across diverse sequencing conditions.

New approaches for molecular analysis continuously open new vistas in molecular medicine. Our lab has built a series of molecular detection techniques based on enzymatic ligation of synthetic oligonucleotides as versatile tools to gain new molecular insights. These fundamental techniques continue to yield new means for specific, high-throughput analyses of nucleic acids and proteins in contexts of relevance for molecular medicine. The combined potential for vast multiplexing and low sample consumption renders the assays described herein attractive as a basis for AI-assisted model building in medicine. Accordingly, this overview is aimed to present a ligase-based molecular toolbox to choose from in addressing present and upcoming analytical needs.

Exploring the dynamical and structural properties of molecular complexes involving DNA is a fundamentally important aspect of understanding many biological processes. Although tools exist for modeling linear DNA and simple complexes, significant challenges remain in generating intricate biomolecular assemblies and incorporating biologically relevant modifications. These limitations restrict the ability to create accurate starting configurations for advanced molecular simulation studies. Here, we introduce MDNA, a molecular modeling toolkit that bridges these gaps by enabling the construction and analysis of complex DNA structures. MDNA provides a versatile solution to generate DNA shapes using a spline-based mapping technique that enables the construction of DNA configurations with arbitrary shapes. Key features include support for (non-)canonical base modifications, such as Watson–Crick–Franklin to Hoogsteen transitions, DNA methylation, and the ability to refine structures using Monte Carlo minimization. The toolkit also provides geometric analysis tools based on rigid body formalism to evaluate DNA structures and trajectories. Together, these features enable users to model and analyze DNA configurations in high detail with a modular Python interface. By integrating structure generation and analysis into a single workflow, MDNA facilitates the study of DNA–protein interactions, supporting new insights into DNA dynamics and molecular simulations.

Single-cell perturbation (Perturb-seq) screens have primarily relied on Cas9 for inducing loss-of-function phenotypes, whereas Cas12a, despite its unique effectiveness for multiplex guide expression, remains underexplored. This may be due to Cas12a’s guide RNA array (pre-crRNA) self-processing activity and the subsequent challenges associated with pre-crRNA sequence recovery during single-cell RNA sequencing library preparation. To overcome the self-processing constraint, we optimized pre-crRNA expression vectors and established a degron-based, enhanced Cas12a system for gene knock-out. As demonstrated across cell types, target genes, and with a minimized guide RNA library, this platform allows for accurate detection of pre-crRNAs and gene editing-induced effects on the transcriptome in single cells. Additionally, we show that HyperLbCas12a outperforms other existing variants for multiplexed gene suppression. While the rapid reversibility of this repressor highlights specific kinetic constraints for degron-based single-cell recording, the system provides a potent, modular tool for contexts requiring tunable, transient silencing. Together, this suite of technologies greatly expands the possibilities for future Perturb-seq efforts and broader application of Cas12a for genetic disruption at scale. Cas12a self-processing of guide RNA transcripts limits its use for Perturb-seq by preventing guide capture. Here, authors develop a chemically degradable Cas12a platform enabling accurate guide assignment in gene knockout experiments and potent multiplexed, transient gene suppression.