We introduce mRNA-GPT, a generative model for end-to-end full-length mRNA sequence design and optimization. Unlike existing approaches that optimize isolated regions, mRNA-GPT jointly optimizes across all three regions (5' UTR, CDS, and 3' UTR) to capture long-range sequence dependencies and cross-region regulatory interactions critical for therapeutic efficacy. The model is pre-trained on 30 million full-length natural mRNA sequences across diverse species and organisms, establishing a robust foundation for sequence generation. We employ Reinforcement Learning (RL), specifically Proximal Policy Optimization (PPO) with oracle-based reward signals, to directly and iteratively optimize target properties, such as half-life and translation efficiency. mRNA-GPT supports flexible generation modes: single regions (UTR or CDS alone), full-length sequences, or generation of any region conditioned on any other region. Through multi-objective optimization, mRNA-GPT achieves Pareto-optimal designs that balance competing properties without sacrificing performance on either objective. mRNA-GPT demonstrates superior design capabilities compared to state-of-the-art methods, achieving enhanced performance in 3' UTR stability optimization, CDS translation rate enhancement, and comprehensive full-length sequence design.

The functional annotation of protein sequences has undergone tremendous progress over recent years, but still too-many protein sequences remain as so-called hypothetical proteins after applying state-of-the-art genome annotation software pipelines. Here, we introduce Baktfold, a new command line software tool for the ultra-sensitive but taxon-independent fast annotation of protein sequences across the microbial tree of life. Baktfold conducts sequential protein structure-based searches against four complementary structure databases. Protein sequences are transformed into Foldseek 3Di tokens via the ProstT5 protein language model and subsequently searched against structure databases via Foldseek. All results are exported in GFF3 and INSDC-compliant flat files as well as comprehensive JSON files facilitating automated downstream analysis 100% interoperable with the popular bacterial annotation tool Bakta. We compared Baktfold's performance in terms of wallclock runtime and functional annotation of hypothetical proteins from various sources including bacterial and archaeal isolates, plasmids, metagenomic-assembled genomes and micro-eukaryotes. When benchmarked on over three hundred thousand species representatives across the prokaryotic tree of life, Baktfold;s median overall bacterial genome annotation rate is 87.8% compared to 72.9% with Bakta, while Baktfold's median bacterial annotation rate of remaining hypothetical proteins is 50.1% (n=290258). For archaea, Baktfold's overall median annotation rate is 71.5% compared to Prokka's 35.8%, with a median archaeal annotation rate of hypothetical proteins of 68.0% (n=14058), making Baktfold the most sensitive automated archaeal annotation method by far. Baktfold is implemented in Python 3 and runs on MacOS and Linux systems. It is freely available under a MIT license at https://github.com/gbouras13/baktfold.

Plasmid conjugation is central to plasmid maintenance and spread among bacteria. Conjugation assays in liquid or on solid agar media are commonly used to quantify plasmid conjugation rates. Plasmids with short, rigid conjugative pili are thought to conjugate more efficiently on surfaces, whereas plasmids encoding long, flexible pili can conjugate efficiently in liquid medium. However, this pattern has not been tested systematically. Here, we perform standardized conjugation assays on a collection of 13 conjugative plasmids belonging to families that play a key role in AMR transmission and encode different conjugative pili types. We confirm that only the plasmids encoding long flexible pili conjugate efficiently in liquid. Furthermore, most transconjugants that arise from liquid assays involving plasmids with short, rigid pili can be attributed to transfer happening after the assay itself, on the surface of selective plates. This effect is amplified when using auxotrophic rather than antibiotic resistance markers, and impacts measures of transfer and defence efficiency. Finally, most of the tested plasmids with short pili had very high conjugation rates on surfaces, suggesting their transfer is mostly limited by physical constraints.

Pyrethroid insecticides are extensively applied owing to their potent insecticidal activity and low mammalian toxicity, yet their hydrophobicity results in persistent environmental residues. Here, we engineered a Bacillus subtilis spore surface display system to anchor Pseudomonas aeruginosa aminopeptidase (PaAps) on the spore coat and evaluated its potential for pyrethroid degradation. Surface-anchored PaAps efficiently degraded various pyrethroids, with β-cypermethrin showing the highest removal. The enzyme exhibited remarkable thermal stability and pH tolerance, with optimal activity at 60 °C and pH 8.0, along with enhanced long-term storage stability. Soil microcosm studies revealed that PaAps not only accelerated β-cypermethrin degradation but also influenced the indigenous microbial community to enhance bioremediation. This study demonstrates a robust and environmentally sustainable biocatalytic strategy for pyrethroid detoxification, with promising applications in environmental remediation and food safety.

Phase variation is viewed as a simple stochastic ON/OFF switch helping bacteria survive unpredictable environments. In minimal-genome pathogens like Mycoplasma bovis, simple sequence repeats (SSRs) introduce frameshifting InDels in key phase-variable genes, such as the Type III restriction-modification mod genes, typically assumed to result in binary expression. This study revisits this assumption using a heterologous E. coli system and single-cell mEGFP-based fluorescence profiling of M. bovis mod1 gene fragment containing SSR to determine if "OFF" states are truly silent. We find that the frameshifted construct remains active, showing low-level, heterogeneous expression in a frame-dependent manner, via frameshift suppression. This creates a range of expression rather than a strict binary switch. These findings suggest that phase-variable SSRs can function with upstream switches to form complex XNOR Boolean logic gates. This demonstrates that sophisticated logic gates can emerge directly from coding sequence architecture enhancing diversity and adaptability to promote evolutionary resilience in compact bacterial genomes.

Intrinsically disordered proteins (IDPs) play an important role in a wide range of biological functions and are linked to several diseases. Due to technical difficulties and the high cost of experimental determination of disorder in proteins, combined with the exponential increase of unannotated protein sequences, the development of computational methods for disorder prediction became an active area of research in the last few decades. In this work, we present emb2dis, a deep learning model that uses protein language models (pLMs) to predict disorder from sequence. The emb2dis tool is a pre-trained model that receives as input a protein sequence, calculates its pLM embedding and passes it to a deep learning model. In contrast to existing approaches, emb2dis integrates informative sequence representations with a novel architecture that combines residual networks (ResNets) and dilated convolutions. This design effectively enlarges the receptive field of the convolution operation, enabling the model to better capture an extended context of each amino acid. At the output, emb2dis assigns a disorder propensity score to each residue in the sequence. The model was evaluated on datasets from the latest CAID3 blind benchmark for disorder prediction, where it achieved first place in the Disorder-PDB category, exhibiting strong performance with high AUC and Fmax scores. Additionally, it ranked among the top ten methods on the Disorder-NOX dataset. We provide a freely available web-demo for emb2dis and a source code repository for local installation. Weblink for the tool: https://sinc.unl.edu.ar/web-demo/emb2dis/

Predicting protein-protein binding affinity from sequence alone remains a bottleneck for antibody optimization, biologics design and large-scale affinity modelling. Structure-based methods achieve high accuracy but cannot scale when complex structures are unavailable. Here we present a framework that reframes affinity prediction as metric learning: two proteins are projected into a shared latent space in which cosine similarity directly correlates with experimental binding affinity, and the protein language model encoder is adapted through parameter-efficient fine-tuning (PEFT). On the PPB-Affinity benchmark, the model achieves Pearson r = 0.89 on a random split, generalises to evolutionarily distant proteins (r = 0.61 at <30% sequence identity) and surpasses structure-based deep learning baselines across biological subgroups, without any three-dimensional input. On the strictly de-overlapped AB-Bind dataset, few-shot adaptation with 30% of assay data (Pearson r = 0.756, RMSE = 0.688) outperforms methods trained on 90% of data; consistent gains are observed across nine diverse AbBiBench deep-mutational-scanning assays with 10-30% labelled variants. Residue-level explainability reveals that the model concentrates importance on interface-localised residues aligned with experimentally validated interaction hotspots across enzyme-inhibitor, and antibody-antigen systems. Together, these results establish a scalable, explainable and data-efficient route to protein-protein binding affinity prediction and therapeutic antibody optimisation from sequence alone.