The compact CRISPR-Cas12f system is promising for AAV-delivered gene therapy, but its application has been constrained by restrictive PAM recognition (e.g., TTTR) and suboptimal editing efficiency. Through bacterial library screening and mammalian cell validation, we engineer evoCas12f, an optimized variant incorporating five key mutations, that dramatically expands PAM recognition to NTNR/NYTR. This advancement reduces median distance between two neighbouring PAM sites to 2 nucleotides in the human genome. It also demonstrates 1.4-fold enhanced activity at TTTR sites compared to wild-type Un1Cas12f1, achieving up to 91% editing efficiency. Remarkably, evoCas12f enables efficient generation of homozygous mutations in F0 generation mice, even at non-canonical PAM sites. We further adapt this system for robust transcriptional activation and precise base editing with a well-defined editing window. As a compact yet highly efficient platform, evoCas12f represents a significant advance in CRISPR technology, enabling multiplexed editing for high-resolution targeting applications and expanding possibilities for therapeutic genome engineering. Compact Un1Cas12f1 enzyme suits AAV delivery but is limited by narrow PAMs and modest activity. Here, authors engineer evoCas12f to recognize broader NTNR/NYTR PAMs and boost editing efficiency up to 91%, enabling efficient multiplex editing, base editing, and gene activation in cells and mice.

Anabaena, a nitrogen-fixing cyanobacterium from the Nostocaceae family, is a promising chassis for sustained space exploration. Using NEKO, this study surveys 2000 publications on Anabaena and summarizes the research, including systems biology, modeling, and potential space applications. Future research should examine its metabolism in space environments, such as microgravity and radiation, and develop bioreactor designs and genetic tools for reliable, self-regulating biomanufacturing systems in these environments.

Sense codon reassignment enables ribosomal incorporation of nonproteinogenic amino acids (npAAs) at any of the 61 sense codons. Because npAAs replace proteinogenic amino acids (pAAs), the total number of available building blocks usually remains limited to 20. To overcome this, we previously introduced “artificial codon box division”, where four-codon boxes (e.g. Val GUN) are split into distinct sets (e.g. GUY and GUG) using in vitro transcribed transfer RNAs (tRNAs) lacking nucleotide modifications. This allows two different amino acids—a pAA and an npAA—to be assigned within the same original box. While we previously demonstrated this by incorporating 23 amino acids, low incorporation efficiency hindered further expansion. Here, we applied our engineered tRNAs, tRNAPro1E2 and tRNAiniP, to the codon box division framework and optimized translation conditions to facilitate multiple npAA incorporations. Consequently, we successfully expanded the genetic code to 32 amino acids, incorporating 11 elongator npAAs and 1 initiator npAA while maintaining all 20 pAAs. Notably, these npAAs include therapeutically significant monomers such as β-amino, d-amino, and N-methyl amino acids, as well as an initiator N-chloroacetyl-d-tyrosine for peptide macrocyclization. This platform offers vast potential for generating diverse macrocyclic peptide libraries with unique chemical entities for drug discovery.

The rich information encoded in cis-regulatory DNA sequences has not been fully exploited for gene function prediction in reverse genetics. Here we show that orthologous cis-regulatory sequences that diverged approximately 160 million years ago share little sequence similarity, yet remarkably retain semantic similarity that can be effectively captured by a deep learning model, PhytoBabel. Although trained solely on orthologous cis-regulatory sequence pairs from 15 angiosperms, PhytoBabel implicitly learned spatio-temporal gene expression patterns, conserved noncoding sequences, semantically similar fragments and phylogenetic relationships among species. Furthermore, PhytoBabel enables the discovery of evolutionarily unrelated but semantically similar cis-regulatory sequences, facilitating the identification of novel genes with functions of interest. As a proof of concept, we identified somatic embryogenesis-related morphogenic regulators in maize that exhibit semantic similarity to known Arabidopsis morphogenic regulators. By bridging the gap in the cis-regulatory sequence → semantics → gene function information chain, PhytoBabel provides a valuable tool for gene function prediction in reverse genetics. PhytoBabel is an AI model designed to capture the semantic similarity of cis-regulatory sequences across phylogenetically distant plant species, thereby establishing an approach for cross-species knowledge transfer and gene function prediction.

Intrinsically disordered proteins and regions (collectively IDRs) are found across all kingdoms of life and have critical roles in virtually every eukaryotic cellular process. IDRs exist in a broad ensemble of structurally distinct conformations. This structural plasticity facilitates diverse molecular recognition and function. Here we combine advances in physics-based force fields with the power of multi-modal generative deep learning to develop STARLING, a framework for rapid generation of accurate IDR ensembles and ensemble-aware representations from sequence. STARLING supports environmental conditioning across ionic strengths and demonstrates proof of concept for the interpolative ability of generative models beyond their training domain. Moreover, we enable ensemble refinement under experimental constraints using a Bayesian maximum-entropy reweighting scheme. Beyond ensemble characterization, STARLING sequence representations can be used in multiple ways. We showcase two examples: first, STARLING lets us perform ensemble-based search for ‘biophysical look-alikes’. Second, we demonstrate how these latent representations can be used to accelerate ensemble-first sequence design from weeks or hours per candidate to seconds, enabling library-scale designs. Together, STARLING dramatically lowers the barrier to the computational interrogation of IDR function through the lens of emergent biophysical properties, complementing bioinformatic protein sequence analysis. We evaluate the accuracy of STARLING against extant experimental data and offer a series of vignettes illustrating how STARLING can enable rapid hypothesis generation for IDR function and aid the interpretation of experimental data. The deep learning model STARLING can generate accurate ensembles of intrinsically disordered regions of proteins using only protein sequence as input.

Biological functions depend on the spatiotemporal distribution of proteins within cells. Key cellular activities such as signal transduction, metabolism, cell cycle and cell death are driven by the interactions of proteins that are localized in multiple cellular compartments. Such multilocalization can even allow protein with identical sequences to display multifunctionality, a phenomenon known as moonlighting. Despite its biological importance, the relationship between protein localization and function remains underexplored. In this Review, we discuss the known mechanisms of protein localization (including RNA transport, role of proteoforms and molecular interactions) and how subcellular localization controls protein function. Proper regulation of protein localization is crucial for specialized cell and tissue functions, including cell differentiation, polarization and the epithelial–mesenchymal transition. Protein mislocalization can also have important roles in pathological processes, such as in cancer, neurodegeneration and autoimmunity. We end with a discussion of current technological and conceptual challenges in the field of subcellular proteomics and spatial biology. Addressing these challenges will allow us to link the dynamic nature of protein localization and function across biological scales and contexts, with great impact on fundamental cell biology and clinical applications. Proteins can localize to multiple cellular compartments and some exhibit distinct functions depending on their location. This Review discusses the mechanisms of protein localization, the control of specialized protein functions through subcellular localization, and how mislocalization is involved in cancer, neurodegeneration and autoimmunity.

DNA superhelicity and transcription are intimately related because changes to DNA topology can influence gene expression and vice versa. Information is transferred through the modulation of local DNA torsional stress, where the expression of one gene may influence the superhelical level of neighboring genes, either promoting or repressing their expression. In this work, we introduce a one-dimensional physical model that simulates supercoiling-mediated regulation. This TORCphysics model takes as input a genome architecture represented either by a plasmid or chromosomal DNA sequence with ends constrained under specific biological conditions and computes the molecule’s output. Our findings demonstrate that the expression profiles of genes are directly influenced by the gene circuit design, including gene location, the positions of topological barriers, promoter sequences, and topoisomerase activity. The novelty that TORCphysics offers is versatility, where users can define distinct activity models for different types of proteins and protein-binding sites. The aim of this research is to establish a flexible framework for developing physical simulations of gene circuits to deepen our comprehension of the intricate mechanisms involved in gene regulation.