Articles

Macrocyclic compounds hold promise as therapeutic agents, yet their structural optimization is hindered by a scarcity of bioactive candidates. Here, the authors present CycleGPT, a generative chemical language model that enhances macrocycle design through innovative transfer learning and sampling strategies, leading to potent JAK2 inhibitors with promising in vivo efficacy.

The calorimetric determination of enthalpies of mixing in multi-component molten salt systems often relies on empirical models that lack physically interpretable parameters. Here, the authors use the molecular interaction volume model (MIVM) to integrate experimentally measured enthalpies and solvation structures from ab initio molecular dynamics simulations to extrapolate excess Gibbs energy and determine the compositional dependence of La³⁺ activity in the LaCl₃-(LiCl-KCl) system.

Nature-inspired catalysts strive to emulate the efficiency and selectivity of enzymes while maintaining the robustness of synthetic catalysts. Here, the authors use bioinformatics to design an eight-amino-acid peptide that self-assembles with copper ions, forming a complex that mimics the laccase enzyme’s active site, highlighting the potential of short peptides in biomimetic catalysis.

Polycyclic aromatic hydrocarbons (PAHs) are key precursors in soot formation, but the reactions leading to clustering in the particulate phase remain under debate. Here, the authors study PAH radical-driven pathways using synchrotron-based and mass-selected vacuum ultraviolet photoelectron spectroscopy and show that early clustering might be driven by π-delocalized radicals, while more localized radicals might contribute to covalent cross-linking during later stages of soot inception.

The measles virus relies on the intrinsically disordered domain of its nucleoprotein, N_TAIL, to bind the polymerase complex responsible for viral transcription and replication, but the role played by disordered regions away from the binding site is not clearly understood. Here, through a combination of experiments and simulations, the authors show that transient and non-local interactions between disordered regions distant in sequence influence the conformational preferences of the binding sites and the folding and availability of its molecular recognition element, affecting viral replication kinetics.

Proteins can act as suitable gas barrier layers in packaging, offering a biodegradable alternative to ethylene vinyl alcohol, however, protein-rich microbial biomass sources for bioplastics remain underexplored. Here, the authors develop bioplastic films and trays from bacterial biomass produced using liquid side streams from the potato processing industry, demonstrating its oxygen barrier properties for packaging applications.

Although spiral growth in noble metals has been observed, its underlying mechanisms remain poorly understood. In this work, the authors demonstrate a high-yield, rapid synthesis of silver nanoplates with spiral morphologies and propose a growth mechanism driven by stochastic mismatches between dendritic arms during planar auto-accelerated oriented assembly, initiated by ligand conformational changes.

Gangliosides play crucial roles in the nervous system, with aberrant metabolism linked to diseases like cancer and neurodegeneration, but tools to visualize and detect gangliosides remain limited and non-specific. Here, the authors introduce MM-JH-2, a dual fluorescent and Raman-active probe, enabling specific ganglioside labeling and differentiation between cells that differ in ganglioside biosynthetic flux.

Diorganozinc reagents (ZnR₂, e.g. R = Et, Ph, C₆F₅) are widely used as Lewis acid catalysts or Lewis base reagents, however, descriptors for predicting the influence of the R substituent are scarce. Here, by using liquid-phase X-ray spectroscopy, the authors have identified the geometric structures of diorganozincs in weakly coordinating solvents and then established Zn-specific descriptors to quantify the properties of their underlying Lewis acidity/basicity.

Glycosylation is a crucial quality attribute of therapeutic proteins, yet existing analytical methods struggle to meet rapid, high-throughput demands. Here, the authors present an optimized glycosylation analysis method using MALDI-TOF-MS, enabling high-throughput analysis with excellent precision and linearity.

Zeolitic imidazolate (ZIF-67), a Co-based metal organic framework, is a promising candidate for Escherichia coli detection, but its application is hampered by poor electrical conductivity. Here, the authors present a Co/Mn-based ZIF-67 that leverages the synergistic action of two transition metals to enhance sensor sensitivity, demonstrating its potential for on-site pathogen monitoring.

The scarcity of experimental training data limits machine learning applications in catalysis research. Here, the authors demonstrate that graph convolutional network models pretrained on a molecular topological index significantly enhance the prediction of catalytic activity, showcasing a promising transfer-learning strategy that leverages self-generated virtual molecular databases.

The ability of the molecular chaperone DNAJB6b to suppress amyloid formation may be linked to its propensity to self-assemble into large micellar-like oligomers, but systematic studies of its solution-phase behaviour are lacking. Here, using a combination of mass photometry and microfluidic diffusional sizing, the authors probe variations in aggregation number distribution and hydrodynamic radius as a function of pH, temperature, ionic strength, or the nature of the anion across the Hofmeister series.

Advancements in organic synthesis often face challenges in optimizing reaction conditions efficiently. Here, the authors develop a fully automated system integrating real-time Fourier-transform infrared spectroscopy and a neural network model, achieving accurate real-time yield predictions for Suzuki–Miyaura cross-coupling reactions.

Mur ligase D (MurD) is a promising antibacterial target, but its dynamics remain underexplored. Here, the authors use ¹⁵N nuclear magnetic resonance to probe apo and bound states of MurD in solution, showing that dynamic events on a time scale of picoseconds to milliseconds have to be considered in the design of structurally novel Mur inhibitors.

N-acyl tyrosines are key biomolecules produced by diverse bacterial phyla, yet their biosynthetic pathways remain largely unknown. Here, the authors identify a pathway for N-acyl dehydrotyrosine biosynthesis in members of the phylum Planctomycetota based on conserved genes encoding a putative adenylyltransferase/cyclase, nitroreductase and the hallmark protein N-acyl amino acid synthase (NasY).

Protein condensates exhibit diverse material properties linked to cellular functions, yet characterizing these properties remains challenging. Here, the authors employ a microfluidic sample deposition and nanometre-resolution mapping technique to characterize the time-dependent material properties in FUS protein condensates, revealing two distinct phase transitions within FUS condensates.

Predicting molecular properties is crucial for drug discovery, yet graph neural networks often fail to capture long-range dependencies. Here, the authors introduce MolGraph-xLSTM, a hierarchical graph-based model that integrates atom-level and motif-level representations with xLSTM to capture long-range intra-molecular dependencies, achieving improved performance across multiple molecular property benchmarks.

Understanding the electronic structure of the actinide series is critical for advancing the nuclear fuel cycle. Here, the authors explore a series of isostructural An(COT^big)₂ (An = Th, U, Np, Pu) complexes with clam-shell geometries, where structural and electronic characterization highlights the impact of f-orbital contributions in actinide bonding.

Mass spectrometry-based CETSAs like PISA face limitations in throughput and scalability due to high sample requirements. Here, the authors integrate PISA with the One-Tip method, reducing cell requirements and enhancing reproducibility, enabling efficient, cost-effective drug-target profiling in a 96-well format.