Most antibiotics are natural compounds or their derivatives, and bacteria have evolved defensive mechanisms to resist them. Many of these mechanisms are still poorly understood or unknown. This study reveals that in Bacillus subtilis, the transcription factor HelD increases resistance to rifampicin by protecting its target, RNA polymerase (RNAP). This protection is mediated by the HelD N-terminal domain that penetrates into RNAP to the close vicinity of the rifampicin binding pocket. Importantly, the bacterium detects low rifampicin levels using a unique regulatory system involving two convergent promoters with finely tuned kinetic properties. In the absence of rifampicin, the stronger antisense promoter inhibits transcription from the sense promoter. In the presence of subinhibitory rifampicin concentration, the antisense promoter is more likely to encounter rifampicin-bound RNAP. This relieves the repression from the sense promoter, increasing its transcription by almost two orders of magnitude, boosting helD expression. A similar two-promoter arrangement also controls the pps gene, which encodes a rifampicin-modifying enzyme. These findings define a widespread bacterial response system sensitive to rifampicin, as this dual-promoter architecture is conserved across many bacterial species and found upstream of genes potentially involved in rifampicin resistance, such as those for hydrolases, transporters, and transferases.

Current base editors act on a maximum of two base substrates and generate limited base conversions or transversions, hindering their applicability for inducing DNA sequence diversity. Here, we engineered a triple base editor (named ACG-BEs) using a fusion of adenine base editor with high A/C catalytic activity and evolved N-methylpurine DNA glycosylase. ACG-BEs enables efficient, multiplexed saturation mutagenesis across adenine (A), cytosine (C), and guanine (G), achieving conversion efficiencies of up to 80.5% for A-to-G/C/T, 75.8% for C-to-T/G/A, and 63.4% for G-to-C/T/A in HEK293T cells. Leveraging ACG-BEs, we identify novel mutations in the HBG1/2 promoter region that confer efficient activation of γ-globin expression in HUDEP-2 cells—a promising advancement for therapeutic strategies targeting hemoglobinopathies. These findings highlight ACG-BEs as a cutting-edge platform for multiplexed saturation mutagenesis, offering broad applications in genetic screening and therapeutic base mutation introduction through enhanced DNA sequence diversity.

RNAs play crucial roles in various important biological functions such as gene regulation and catalysis. The functions of RNAs are generally coupled to their structures as well as to the stability of the structure, which can strongly depend on ionic conditions. However, it is still a challenge to make reliable predictions for the structures and stability of RNAs in ion solutions. In this work, we developed a coarse-grained method involving temperature and ion effects for simulating RNA folding and three-dimensional (3D) structure prediction, named TiRNA. Extensive tests demonstrate that TiRNA can make successful predictions for 3D structures of RNAs, including pseudoknots and multi-way junctions, and for thermal stability of RNAs in ion solutions solely from sequences, as compared with the top existing methods. Moreover, TiRNA can also make reliable predictions for 3D structures and stability of RNAs in ion solutions based on inputting secondary structures.

Increasing quantities of genomic data present an opportunity for uncovering functional and mechanistic insights. However, given the complexity and scale of the data, sophisticated analytical frameworks and integrative approaches are required. In addition, the unique modeling requirements of different data modalities and the large number of disease states, tissue types and developmental profiles pose complex challenges. Current analysis methods are often computationally expensive and technically challenging for the end user to apply. There is a need for comprehensive resources that bridge the gap between raw data and biological insights, allowing researchers at all levels of computational fluency to bring large data compendia and cutting-edge analytical approaches to bear on their questions of interest. Here, we present HumanBase (https://humanbase.io), an interactive web-based platform that uses AI to integrate and interpret large-scale human genomic data.

Here, we present an efficient PE system in rice, developed using a combination of engineered pegRNA (epegRNA) to enhance stability and prevent degradation of the 3′ extension, and paired pegRNAs to target different DNA strands, synthesizing edited DNA sequences simultaneously by RT of each template. Using this approach, we observed that undesired scaffold-derived mutations occurred more frequently in regenerated rice plants than indels resulting from DNA double-strand break (DSB) repair. To improve the fidelity of the precise edits, we next tested a PE system using paired epegRNA-HFs (PE with paired epegRNA-HFs). Our results show that PE with paired epegRNA-HFs in rice avoids the introduction of undesired scaffold-derived byproducts into the target locus, resulting in the establishment of an efficient and accurate PE system in plant cells.

Fecal coprolites preserve ancient microbiomes and are a potential source of extinct but highly efficacious antimicrobial peptides (AMPs). Here, we develop AMPLiT (AMP Lightweight Identification Tool), an efficient tool deployable to portable hardware for AMP screening in metagenomic datasets. AMPLiT demonstrates AUPRC performances of 0.9486 ± 0.0003 and reasonable overall training time of 3200 ± 53 s. By computationally utilizing AMPLiT, we analyze seven ancient human coprolite metagenomes, identifying 160 AMP candidates. Of 40 representative peptides synthesized, 36 (90%) peptides demonstrate measurable antimicrobial activity at 100 μM or less in vitro. Strikingly, approximately two-thirds of these peptides are sourced from Segatella copri, a dominant ancient gut commensal that is conspicuously underrepresented in modern populations, particularly those with Westernized lifestyles. Representative S. copri-derived AMPs exhibit disruptions against membranes of pathogenic bacteria, coupled with low cytotoxicity and hemolytic risk. In vivo, lead peptides demonstrate potent antibacterial and wound-healing efficacy comparable to traditional antibiotics, especially in combating gram-positive pathogens. Our findings highlight the ancient gut microbiomes as sources of novel AMPs, offering valuable insights into the historical role of S. copri in human health and its decline in contemporary populations. Here, the authors develop AMPLiT a tool for screening antimicrobial peptides in metagenomic datasets, and apply it to human coprolite metagenomes, finding that Segatella copri, an ancient prevalent human gut bacterium declined in modern populations, harbors unexplored antimicrobial reservoir, offering an alternative approach against modern pathogenic infections.

CRISPR-Cas12a holds substantial promise for molecular diagnostics, yet its rapid and uncontrolled activation often results in background leakage and disrupts the coordination of upstream reaction modules. Here, we established a steric-regulation framework that enables predictable tuning of Cas12a trans-cleavage kinetics through rationally engineered extensions on split activators. Systematic analysis of extension orientation, length, and hybridization state revealed quantitative and direction-dependent rules governing steric control of activator assembly and Cas12a activation. Guided by these insights, we integrated the sterically regulated split activator into an entropy-driven DNA circuit to construct a fully one-pot cascaded detection system. The engineered steric barriers effectively suppressed premature activation and established precise kinetic matching between the DNA circuit and Cas12a. The resulting platform achieved a detection limit of 1.24 pM for microRNA-21 and demonstrated high fidelity. This work defines a predictable steric-gating mechanism for Cas12a activation and delivers a nucleic-acid-only regulatory module that can be incorporated into diverse CRISPR architectures, supporting the development of robust, leakage-resistant one-pot diagnostic systems.

Enzymes are powerful and sustainable catalysts, but their widespread application is limited by the difficulty of identifying functional starting points for optimization, creating a major bottleneck in early- stage biocatalyst discovery. Designing libraries of such starting enzymes remains particularly challenging. Here, we use the GenSLM protein language model to generate novel β-subunit of tryptophan synthase (TrpB) enzymes that express in Escherichia coli and are both stable and catalytically active. Many generated TrpBs also display significant substrate promiscuity, outperforming their natural counterparts on non-native substrates. Some even surpass laboratory-evolved TrpBs. Comparison of the most-active and most-promiscuous generated TrpB to its closest natural homolog confirms that the enhanced versatility is absent from the natural enzyme, highlighting the creative potential of generative models. These results demonstrate that the generated TrpBs not only preserve natural structure and function but also acquire non-natural properties, establishing generative models as powerful tools for biocatalyst discovery and engineering. Enzymes are highly selective and sustainable catalysts for chemical synthesis, but their optimization is often limited by the difficulty of identifying functional starting points. This study shows that using the GenSLM protein language model to design TrpB variants can yield stable, active enzymes with broad substrate promiscuity, outperforming natural and evolved counterparts and demonstrating the potential of generative models to accelerate biocatalyst discovery.

Automated free energy calculations for the prediction of binding free energies of congeneric series of ligands to a protein target are growing in popularity, but building reliable initial binding poses for the ligands is challenging. Here, we introduce the open-source FEgrow workflow for building user-defined congeneric series of ligands in protein binding pockets for input to free energy calculations. For a given ligand core and receptor structure, FEgrow enumerates and optimizes the bioactive conformations of the grown functional group(s), making use of hybrid machine learning/molecular mechanics potential energy functions where possible. Low energy structures are optionally scored using the gnina convolutional neural network scoring function, and output for more rigorous protein–ligand binding free energy predictions. We illustrate use of the workflow by building and scoring binding poses for ten congeneric series of ligands bound to targets from a standard, high quality dataset of protein–ligand complexes. Furthermore, we build a set of 13 inhibitors of the SARS-CoV-2 main protease from the literature, and use free energy calculations to retrospectively compute their relative binding free energies. FEgrow is freely available at https://github.com/cole-group/FEgrow , along with a tutorial. Automated free energy calculations for the prediction of binding free energies of ligands to a protein target are gaining importance for drug discovery, but building reliable initial binding poses for the ligands is challenging. Here, the authors introduce an open-source workflow for building user-defined congeneric series of ligands in protein binding pockets for input to free energy calculations.