Articles in 2025

Here, the authors generated and analyzed run-on sequencing data to observe transcription in species across the tree of life to uncover the origins of the promoter-proximal pause.

We review 2025 and discuss some of the foremost initiatives developed at the journal. We also look back at discoveries we have been proud to publish.

The AlphaFold database, released in 2022, modeled UniProt sequences from April 2021 and now provides 200 million predicted protein structures. Of the 20,504 full-length predicted human structures, 631 entries conflict with the June 2025 UniProt release. Similar conflicts across species highlight how bioinformatics resources can rapidly age.

Zeng et al. applied single-particle cryo-electron microscopy to native samples isolated from the human parasite Toxoplasma gondii, determining multiple structures of key components of the conoid, a cone-shaped organelle essential for host cell invasion.

Here the authors present an artificial-intelligence-based automated method for improved protein structure determination from cryo-EM density maps by efficiently integrating map information and structure prediction.

Riechmann et al. uncover structural features governing ubiquitin transfer from ubiquitin-activating E1 enzymes UBA1 and UBA6 to specific E2 ubiquitin-conjugating enzymes, revealing a hierarchy of E2 activity with cognate E1s.

Mitochondrial translational activators (TAs) facilitate transcript-specific translation. Using selective ribosome profiling and cryo-electron microscopy, the authors show that TAs bind to the 5′ untranslated region of their target transcript to position mitoribosomes for initiation.

This Review summarizes the various functions of the exon junction complex in RNA splicing and beyond, to influence gene regulation.

Vidmar et al. use cryo-EM to reveal how bacterial RNA polymerase (RNAP) and topoisomerase I (TopoI) cooperate. TopoI switches conformation, senses DNA supercoils near RNAP and relaxes them. Mutations disrupting this process alter bacterial motility and operon polarity.

Cells store excess fat in lipid droplets to avoid lipotoxicity and maintain homeostasis. We identified an autophagy-independent role for the autophagy lipid transfer protein ATG2A in helping direct lipids to growing lipid droplets and promoting recruitment of the enzyme DGAT2. This coordination enhances triglyceride storage, protects the endoplasmic reticulum from lipid overload and limits the misrouting of lipids into other metabolic pathways.

By reconstituting and visualizing mammalian transcription elongation at the single-molecule level, Wang et al. dissected the effects of individual elongation factors on the speed of RNA polymerase II, which is found to operate as a multi-gear molecular machine.

Rocha-Roa et al. identify TMEM170 proteins as endoplasmic reticulum lipid scramblases that partner with bridge-like lipid transfer proteins BLTP1/Csf1 proteins to enable bulk lipid transport between organelles.

Carty et al. identify the H3K9 methyltransferases that restrict the size and position of the centromere protein A chromatin domain, maintaining functional centromeres.

Artificial intelligence (AI) is advancing genome editing, from predictive modeling to generative design. Emerging generative AI tools such as RFdiffusion, AlphaFold 3 and ESM now facilitate the de novo design of linkers, inhibitors and enzymes. We highlight work where AI-driven design is used to enhance the precision of mitochondrial cytosine base editors.

Mi et al. use de novo protein design to address bystander and off-target editing in base editing, resulting in a highly precise mitochondrial cytosine base editor that is valuable for studying and treating mitochondrial diseases.

Elhan et al. show that ATG2A acts with DGAT2, the enzyme producing triacylglycerol (TAG), in lipid droplet growth. By delivering diacylglycerol to lipid droplets, ATG2A not only fuels TAG production but also promotes the recruitment of DGAT2 to droplet surfaces.

Baretić and Missoury et al. identify vertebrate proteins FAM118B and FAM118A as sirtuins similar to bacterial antiphage enzymes and show that FAM118A/B processing of NAD involves head-to-tail filament formation and a partnership between the two paralogs.

Inspired by the advance in protein structure prediction, CryoAtom builds protein models directly from cryogenic electron microscopy maps, producing more complete models, reducing the resolution requirement and accelerating modeling.