Developing short, stable, and potent antimicrobial peptides is a promising strategy to combat antibiotic resistance and persistence. We present CAMPER (Constraint-driven AMP Engineering with Ranking), a mechanistic artificial intelligence framework that integrates machine learning with biophysical ranking to prioritize membrane-targeting peptides effective against persister and biofilm forms of methicillin-resistant Staphylococcus aureus. We apply CAMPER to identify WP-CAMPER1 (12mer) that kills S. aureus MW2 at a minimal inhibitory concentration of 4 µg/mL. A 2% topical WP-CAMPER1 formulation reduces S. aureus MW2 burden by 2.5 log10 (p < 0.0002) in a murine prophylactic skin infection model, while its D-enantiomer, WP-CAMPER1-d, achieves 1.37 log10 (p < 0.0001) reduction in an established biofilm infection model. Single-cell analysis using a high-throughput microfluidic system shows that WP-CAMPER1-d reduces exponential-phase persisters of S. aureus USA300, and, in a deep-seated murine thigh infection model, decreases stationary-phase S. aureus MW2 persisters by 1.6 log10 (p < 0.0001). This study introduces CAMPER, a mechanistic artificial intelligence platform for designing antimicrobial peptides targeting MRSA. CAMPER identified a stable peptide that eradicates MRSA biofilms and persister cells and was active in mouse infection models.

Gene regulatory networks (GRNs) explain how the genome controls cellular behavior and tissue morphogenesis, serving to connect molecular mechanism to functional output. Single-cell technologies now provide descriptions of these networks with unprecedented detail, but this advance has also revealed gene regulatory systems that are too complex for our existing conceptual frameworks. GRNs, which should provide mechanistic explanations, are increasingly reduced to statistical correlations — ‘hairballs’ that fail to capture molecular causation. Here, we explore why this dilemma exists and propose a path forward. We argue that methods in ‘representation learning’ can be used to model GRNs, without needing to capture every molecular detail. For this framework, we advocate three linked principles: models must be inherently mechanistic, with structures grounded in cellular and evolutionary biology; molecular principles and constraints must be used to reduce the solution space for learning GRN models; and more sophisticated forms of experimental perturbation and synthetic biological engineering are needed to train models and test predictions. By reimagining GRNs through these principles, we can bridge the gap from data abundance to new conceptual understanding. In this Perspective, Maizels and Briscoe discuss the limitations of current models of gene regulatory networks and outline solutions to harness data abundance without compromising explanatory power.

Within lignocellulosic biomass, xylose is the second most abundant sugar after glucose. As a renewable and sustainable substrate, it is gaining attention as a feedstock for microbial bioprocesses. In this study, we demonstrated the co-production of polyhydroxybutyrate (PHB) and violacein from xylose. We initially confirmed the feasibility of co-production through genome-scale metabolic simulations, followed by optimization using a hybrid expression system that combines a conventional tac promoter and synthetic promoter-ribosome-binding site-terminator (semi-endo PRT) elements in dual plasmids. Additionally, we assessed the antimicrobial activity of violacein against type I methanotrophs.  Recombinant E. coli DH5α harboring a hybrid system produced 111.3 ± 19.7 and 0.88 ± 0.23 mg/L of PHB and violacein, respectively, in M9 using xylose as the sole carbon source, without tryptophan supplementation. Using the synthetic PRT system, 528.9 ± 104 mg/g dry cell weight (DCW) of PHB was obtained. Additionally, violacein inhibits the growth of Methylomicrobium alcaliphilum 20Z at 9 µg/mL in nitrate mineral salt medium containing methanol as the sole carbon source. The use of lignocellulose-derived sugars for co-production offers an environmentally sustainable bio-manufacturing approach that contributes to greenhouse gas mitigation and supports the transition toward a circular bioeconomy.

Protein vesicles are spherical, hollow structures made entirely of folded proteins, fusion proteins, or polypeptides. Their intrinsic biocompatibility, nontoxicity, structural tunability, and cargo-loading capacity make them promising candidates for diverse biomedical applications. Although diverse forms of protein-based carriers have long been employed in drug, gene, and vaccine delivery, as well as in artificial antigen-presenting cells, vesicle architectures provide distinct advantages over free proteins, including enhanced stability, targeted delivery, and controlled release. We summarize recent advances in engineering protein vesicles and assess their current status within the broader landscape of synthetic vesicles in biomedicine. By comparing protein vesicles with liposomes, polymersomes, and virus-like particles, we highlight the limitations of conventional systems and underscore the unique benefits of protein-based assemblies. We further examine the emerging applications of protein vesicles in therapeutic delivery, diagnostics, and immunotherapy and discuss future directions needed to advance protein vesicle technologies toward clinical translation.

Proteins are essential components of all living organisms and play a critical role in cellular survival. They have a broad range of applications, from clinical treatments to material engineering. This versatility has spurred the development of protein design, with amino acid sequence design being a crucial step in the process. Recent advancements in deep generative models have shown promise for protein sequence design. However, the scarcity of functional protein sequence data for certain types can hinder the training of these models, which often require large datasets. To address this challenge, we propose a hierarchical model named ProteinRG that can generate functional protein sequences using relatively small datasets. ProteinRG begins by generating a representation of a protein sequence, leveraging existing large protein sequence models, before producing a functional protein sequence. We have tested our model on various functional protein sequences and evaluated the results from three perspectives: multiple sequence alignment, t-SNE distribution analysis, and 3D structure prediction. The findings indicate that our generated protein sequences maintain both similarity to the original sequences and consistency with the desired functions. Moreover, our model demonstrates superior performance compared twith other generative models for protein sequence generation.

Human protein-based biomaterials facilitate the development of preclinical models that integrate predictive value and ethical responsibility. Metatissue uses decellularized extracellular matrix from placenta samples and human blood to create tunable, xeno-free substrates for 3D cell culture, minimizing the reliance on animal models and accelerating translational research.

Nucleic acid degradation is a common strategy for prokaryotic antiphage systems, as exemplified by the CRISPR-Cas system. The PD-(D/E)-XK nucleases constitute a widely distributed family in these defenses. Notably, most members exhibit a single nuclease domain, while variants containing dual nuclease domains within a single polypeptide remain underexplored, and their molecular mechanisms largely obscure. Here, we biochemically and functionally study a single-protein system containing an uncharacterized PD-(D/E)-XK defense protein (Upx). As revealed by single-particle electron cryo-microscopy (cryo-EM) structure, the C-terminal domain (CTD) harboring the conserved PD-(D/E)XK catalytic core is buttressed by the N-terminal domain (NTD) and the middle domain (MD). Functional assays demonstrate that the nucleic acid binding capability of the CTD is enhanced by the MD. The NTD also displays a noncanonical, basal exonuclease activity that is auto-inhibited by MD. IP-MS experiments identify Upx-interacting phage proteins, and substrate profiling defines its physiological preferences, collectively pointing to its potential physiological targets. Notably, the phage protein gp16 was found to relieve MD-mediated inhibition of the NTD, suggesting a virus-triggered mechanism for activating Upx’s dual nuclease activity. Together, these findings establish Upx as a single-protein dual-nuclease anti-phage system, expanding our understanding of bacterial immunity and informing antiviral strategy development. Bacteria employ diverse nucleases to defend against viruses. Here, authors characterize Upx, a single-protein, dual-nuclease system, and reveal that a specific phage protein relieves auto-inhibition to activate its antiviral function.

Disordered proteins play essential roles in myriad cellular processes, yet their structural characterization remains a major challenge due to their dynamic and heterogeneous nature. Here we present a community-driven initiative to address this problem by advocating a unified framework for determining conformational ensembles of disordered proteins. Our aim is to integrate state-of-the-art experimental techniques with advanced computational methods, including knowledge-based sampling, enhanced molecular dynamics and machine learning models. The modular framework comprises three interconnected components: experimental data acquisition, computational ensemble generation and validation. The systematic development of this framework will ensure the accurate and reproducible determination of conformational ensembles of disordered proteins. We highlight the open challenges necessary to achieve this goal, including force-field accuracy, efficient sampling, and environmental dependence, advocating for collaborative benchmarking and standardized protocols. This Perspective establishes a comprehensive and practical framework to guide intrinsically disordered protein (IDP) ensemble determination, benchmarking and interpretation, as well as proposes a roadmap for IDP ensemble determination, uncertainty quantification and actionable benchmarking strategies.

All organisms fuel and build themselves through their energy metabolism. While classic biochemistry conceptualizes the fluxes of energy and matter, our understanding of how energy metabolism works in vivo contains major gaps. One reason is that energy metabolism spans multiple scales, from enzymes to organs, and shifts dynamically at environmental and developmental transitions, resulting in a degree of complexity that is presently impossible to capture. Genetically encoded fluorescent biosensors have started to bridge several critical gaps by enabling live monitoring of metabolites across scales. Recently, several paradigms of energy metabolism have started to shift, driven by the expansion of biosensing tools. This review explores advancements in our understanding of plant energy metabolism driven by fluorescent protein biosensing, highlights emerging concepts and open questions, and discusses how available tools, and much-needed future innovations, can unlock the potential of biosensing toward understanding in vivo plant energy metabolism and its effective modification.

Protein–protein interactions (PPIs) are central to cellular signaling and regulation, and their dysregulation underlies many diseases. Predicting the impact of mutations on PPI stability, quantified as ΔΔG, is essential for understanding disease mechanisms and guiding protein engineering. Here, we first present MutPPI, a graph-based deep-learning model that encodes full-residue structural features of protein–protein complexes and employs a shared GIN-GAT feature extractor for wild-type and mutant complexes. MutPPI outperforms 12 existing methods on an antibody–antigen single-point mutation dataset (S645). By integrating evolutionary information from protein language models, we further develop MutPPI-plus, achieving enhanced predictive performance. Second, we proposed a mutation-path-based data augmentation strategy, which enriches input modalities and improves generalization of both MutPPI and MutPPI-plus. After data augmentation, MutPPI-plus demonstrates state-of-the-art performance on S645 and three additional multi-point mutation datasets (SM_ZEMu, SM595, SM1124), substantially surpassing DDMut-PPI. Our analyses highlight the benefits of the multimodal framework and the physically informed data augmentation method. Together, these results provide a versatile computational tool for accurate ΔΔG prediction, advancing rational protein design.