US-align: universal structure alignments of proteins, nucleic acids, and macromolecular complexes

SidF, a dual substrate N5-acetyl-N5-hydroxy-L-ornithine transacetylase involved in Aspergillus fumigatus siderophore biosynthesis

Siderophore-mediated iron acquisition is essential for the virulence of Aspergillus fumigatus, a fungus causing life-threatening aspergillosis. Drugs targeting the siderophore biosynthetic pathway could help improve disease management. The transacetylases SidF and SidL generate intermediates for different siderophores in A. fumigatus. A. fumigatus has a yet unidentified transacetylase that complements SidL during iron deficiency in SidL-lacking mutants. We present the first X-ray structure of SidF, revealing a two-domain architecture with tetrameric assembly. The N-terminal domain contributes to protein solubility and oligomerization, while the C-terminal domain containing the GCN5-related N-acetyltransferase (GNAT) motif is crucial for the enzymatic activity and mediates oligomer formation. Notably, AlphaFold modelling demonstrates structural similarity between SidF and SidL. Enzymatic assays showed that SidF can utilize acetyl-CoA as a donor, previously thought to be a substrate of SidL but not SidF, and selectively uses N5-hydroxy-L-ornithine as an acceptor. This study elucidates the structure of SidF and reveals its role in siderophore biosynthesis. We propose SidF as the unknown transacetylase complementing SidL activity, highlighting its central role in A. fumigatus siderophore biosynthesis. Investigation of this uncharacterized GNAT protein enhances our understanding of fungal virulence and holds promise for its potential application in developing antifungal therapies.

Accurate prediction of virulence factors using pre-train protein language model and ensemble learning

BACKGROUND

As bacterial pathogens develop increasing resistance to antibiotics, strategies targeting virulence factors (VFs) have emerged as a promising and effective approach for treating bacterial infections. Existing methods mainly relied on sequence similarity, and remote homology relationships cannot be discovered by sequence analysis alone.

RESULTS

To address this limitation, we developed a protein language model and ensemble learning approach for VF identification (PLMVF). Specifically, we extracted features from protein sequences using ESM-2 and their three-dimensional (3D) structures using ESMFold. We calculated the true TM-score of the proteins based on their 3D structures and trained a TM-predictor model to predict structural similarity, thereby capturing hidden remote homology information within the sequences. Subsequently, we concatenated the sequence-level features extracted by ESM-2 with the predicted TM-score features to form a comprehensive feature set for prediction. Extensive experimental validation demonstrated that PLMVF achieved an accuracy (ACC) of 86.1%, significantly outperforming existing models across multiple evaluation metrics. This study provided an ideal tool for identifying novel targets in the development of anti-virulence therapies, offering promise for the effective prevention and control of pathogenic bacterial infections.

CONCLUSIONS

The proposed PLMVF model offers an efficient computational approach for VF identification.

Conserved Local Structural Motifs in Glycoside Hydrolase Families Facilitate the Discovery of Functional Enzymes

Glycoside hydrolases (GHs) are vital for natural glycoside biotransformation, especially in enhancing the pharmacological effects of natural products like ginsenosides. In this study, we collected 67 microbial-derived ginsenoside-hydrolyzing enzymes from nine GH families. Despite differences in global structures, the key residues surrounding substrate binding in GH1 and GH3 exhibit conserved structural motifs. Leveraging these motifs, five GH genes from Cellulosimicrobium were cloned, and three enzymes (Cbgl496, Cbgl516, Cbgl766) were characterized. Experimental results demonstrated that Cbgl516, Cbgl766, and Cbgl841 specifically catalyzed the hydrolysis of the β(1-6) glycosidic bond in the C-20 sugar chain of ginsenoside Rb1 to yield Rd. Cbgl496 selectively catalyzed the hydrolysis of β(1-2) glycosidic bonds in the oligosaccharide chains at the C-3 position of ginsenosides Rb1, Rb2, Rb3, and Rc, thereby directionally producing the minor ginsenosides Gy XVII, Compound O, Compound Mx1, and Compound Mc1. Structural analysis of 109,994 GH1/GH3 models from AlphaFold database revealed conserved residues across various organisms, emphasizing evolutionary conservation in the 3D structure of the catalytic core region despite sequence diversity. This study underscores the importance of conserved local structural motifs in GHs, offering insights for functional enzyme screening and understanding enzyme diversity and industrial applications.

DEMO-EMol: modeling protein-nucleic acid complex structures from cryo-EM maps by coupling chain assembly with map segmentation

Atomic structure modeling is a crucial step in determining the structures of protein complexes using cryo-electron microscopy (cryo-EM). This work introduces DEMO-EMol, an improved server that integrates deep learning-based map segmentation and chain fitting to accurately assemble protein-nucleic acid (NA) complex structures from cryo-EM density maps. Starting from a density map and independently modeled chain structures, DEMO-EMol first segments protein and NA regions from the density map using deep learning. The overall complex is then assembled by fitting protein and NA chain models into their respective segmented maps, followed by domain-level fitting and optimization for protein chains. The output of DEMO-EMol includes the final assembled complex model along with overall and residue-level quality assessments. DEMO-EMol was evaluated on a comprehensive benchmark set of cryo-EM maps with resolutions ranging from 1.96 to 12.77 Å, and the results demonstrated its superior performance over the state-of-the-art methods for both protein-NA and protein-protein complex modeling. The DEMO-EMol web server is freely accessible at https://zhanggroup.org/DEMO-EMol/.

A novel TRKB-activating internal tandem duplication characterizes a new mechanism of receptor tyrosine kinase activation

Precision medicine programs like the Zero Childhood Cancer Program perform comprehensive molecular analysis of patient tumors, enabling detection of novel structural variants that may be cryptic to standard techniques. Identification of these variants can impact individual patient treatment, and beyond this establish new mechanisms of oncogenic activation. We have identified a novel internal tandem duplication (ITD) in the receptor tyrosine kinase (RTK), NTRK2, in a patient with FOXR2-activated CNS neuroblastoma. The ITD spans exons 10-13 of NTRK2 encoding the transmembrane domain. NTRK2 ITD is transforming and sensitive to TRK inhibition. In silico structural predictions suggested the duplication of an alpha-helix region and juxtaposed tyrosine residues that play a role in facilitating autophosphorylation. Consistent with this, mutation of these residues inhibited cellular transformation. This is the first report of an ITD spanning the transmembrane domain of an RTK, characterizing an additional mechanism by which RTKs are activated in cancer.

Construction of safer hirudin-derived peptides with enhanced anticoagulant properties based on C-terminal active residue adaptation and modification

INTRODUCTION

Hirudin exerts anticoagulant effects by inhibiting the binding and catalytic activity of thrombin to fibrinogen. However, its rigid N-terminal region irreversibly occupies the thrombin catalytic center, raising concerns about potential bleeding.

OBJECTIVES

In this study, a novel lead compound, WPHVC_V1, which is based on the competitive binding mechanism of the hirudin variant WPHV_C, was developed and validated for in vitro and in vivo activity and safety.

METHODS

Saturation mutagenesis, molecular dynamics simulations and mutant protein activity assays were used to elucidate the competitive anticoagulant mechanism between WPHV_C and thrombin. Next, a recombinantly expressed tyrosylprotein sulfotransferase was used to modify and confirm the sulfation site on the C-terminal tyrosine of hirudin. Finally, a multisite aromatic amino acid mutation strategy was implemented to design and synthesize the lead anticoagulant, WPHVC_V1.

RESULTS

The acidic amino acid cluster in WPHV_C formed strong electrostatic interactions with the positively charged thrombin exosite I, blocking fibrinogen binding. The introduction of aromatic amino acids further stabilized the complex through π-π stacking and π-cation interactions. For example, mutation of 13E to A decreased the free energy of dissociation (ΔG) from 19.27 to 10.93 kcal·mol-1 and shortened the thrombin time (TT) from 42.00 s to 30.94 s, whereas mutation of 26 K to W increased the ΔG to 24.70 kcal·mol-1 and prolonged TT to 51.92 s. In addition, the aromatic effect of 20Y, combined with sulfation, synergistically enhanced binding. Based on these findings, the newly designed WPHVC_V1 showed a ΔG of 37.24 kcal·mol-1 and, at 0.1 mg/ml, increased TT/APTT/PT from 41.72/14.38/15.86 s (WPHV_C) to 62.08/23.38/22.22 s. In in vivo studies, WPHVC_V1 achieved tail thrombus inhibition in the mouse tail by reducing the length of the thrombus from 3.562 cm in CK to 1.853 cm (1.729 cm for sodium heparin and 2.530 cm for WPHV_C), completely inhibited thrombus formation in a carotid artery model and reduced tail bleeding time by 35.2 s compared with heparin sodium. Safety evaluations revealed that WPHVC_V1 did not cause hemolysis, had no significant effect on blood pressure or cause pathological changes in major organs.

CONCLUSION

These findings provide an initial foundation and sequence reference for the development of safe and effective anticoagulant drugs with potential for clinical translation.

Mirror substrates specificity of a 2, 3-dihydroxypropanesulfonate degrading enzyme in sulfate-reducing bacteria

Ubiquitous R- and S-enantiomers of 2,3-dihydroxypropanesulfonate (DHPS), organic sulfur compounds produced by photosynthetic organisms, serve as common nutrient and energy sources for specific bacteria. While most known DHPS-degrading enzymes exhibit enantioselectivity, this study introduces a unique dehydrogenase, DhpA from the sulfate-reducing bacterium Desulfovibrio sp. DF1, capable of efficiently metabolizing both R- and S-DHPS to 3-sulfolactaldehyde (SLA). The crystal structure of DhpA reveals a conserved binding pocket that recognizes the sulfonate group of DHPS through interactions with Lys123, Ser174, and Asn175. The catalytic mechanism of the enzyme involves the oxidation of the C3-OH group of both enantiomers, facilitated by the Lys171. The mutation of Lys171 significantly diminishes activity, confirming its critical role in catalysis. Based on biochemical and genetic analyses, this study proposes a chiral DHPS degradation pathway in bacteria. This study reveals the unique enantiomeric selectivity of DhpA, expanding our understanding of the bacterial metabolism of chiral molecules.

Generation of a compound heterozygous ABCA4 rat model with pathological features of STGD1

The ABCA4 protein plays an essential role in mammalian vision, ensuring the correct localization of all-trans-retinal within the visual cycle. Mutations in the ABCA4 gene are responsible for the juvenile maculopathy, Stargardt disease (STGD1). We investigated the most common variant underlying STGD1 phenotype in a rat model carrying the ortholog to the human c.5882G > A/p.(Gly1961Glu) (G1961E) in ABCA4. While the pathogenicity of this variant has recently been questioned, we examine here whether the ortholog rat variant is associated with vitamin A toxicity in the retina. By crossing the rat line with a rat line deficient in ABCA4 protein, we reveal a more pathogenic phenotype in line with compound heterozygosity, making the model suitable for testing of gene, cell and pharmacological therapies.

Prediction and Evaluation of Coronavirus and Human Protein-Protein Interactions Integrating Five Different Computational Methods

The high lethality and infectiousness of coronaviruses, particularly SARS-Cov-2, pose a significant threat to human society. Understanding coronaviruses, especially the interactions between these viruses and humans, is crucial for mitigating the coronavirus pandemic. In this study, we conducted a comprehensive comparison and evaluation of five prevalent computational methods: interolog mapping, domain-domain interaction methodology, domain-motif interaction methodology, structure-based approaches, and machine learning techniques. These methods were assessed using unbiased datasets that include C1, C2h, C2v, and C3 test sets. Ultimately, we integrated these five methodologies into a unified model for predicting protein-protein interactions (PPIs) between coronaviruses and human proteins. Our final model demonstrates relatively better performance, particularly with the C2v and C3 test sets, which are frequently used datasets in practical applications. Based on this model, we further established a high-confidence PPI network between coronaviruses and humans, consisting of 18,012 interactions between 3843 human proteins and 129 coronavirus proteins. The reliability of our predictions was further validated through the current knowledge framework and network analysis. This study is anticipated to enhance mechanistic understanding of the coronavirus-human relationship a while facilitating the rediscovery of antiviral drug targets. The source codes and datasets are accessible at https://github.com/covhppilab/CoVHPPI.

Identifying RNA-small Molecule Binding Sites Using Geometric Deep Learning with Language Models

RNAs are emerging as promising therapeutic targets, yet identifying small molecules that bind to them remains a significant challenge in drug discovery. This underscores the crucial role of computational modeling in predicting RNA-small molecule binding sites. However, accurate and efficient computational methods for identifying these interactions are still lacking. Recently, advances in large language models (LLMs), previously successful in DNA and protein research, have spurred the development of RNA-specific LLMs. These models leverage vast unlabeled RNA sequences to autonomously learn semantic representations with the goal of enhancing downstream tasks, particularly those constrained by limited annotated data. Here, we develop RNABind, an embedding-informed geometric deep learning framework to detect RNA-small molecule binding sites from RNA structures. RNABind integrates RNA LLMs into advanced geometric deep learning networks, which encodes both RNA sequence and structure information. To evaluate RNABind, we first compile the largest RNA-small molecule interaction dataset from the entire multi-chain complex structure instead of single-chain RNAs. Extensive experiments demonstrate that RNABind outperforms existing state-of-the-art methods. Besides, we conduct an extensive experimental evaluation of eight pre-trained RNA LLMs, assessing their performance on the binding site prediction task within a unified experimental protocol. In summary, RNABind provides a powerful tool on exploring RNA-small molecule binding site prediction, which paves the way for future innovations in the RNA-targeted drug discovery.

Using AlphaFold-Multimer to study novel protein-protein interactions of predation essential hypothetical proteins in Bdellovibrio

Bdellovibrio bacteriovorus is the most studied member of a group of small motile Gram-negative bacteria called Bdellovibrio and Like Organisms (BALOs). B. bacteriovorus can prey on Gram-negative bacteria, including multi-drug resistant pathogens, and has been proposed as an alternative to antibiotics. Although the life cycle of B. bacteriovorus is well characterized, some molecular aspects of B. bacteriovorus-prey interaction are poorly understood. Hypothetical proteins with unestablished functions have been implicated in B. bacteriovorus predation by many studies. Our approach to characterize these proteins employing Alphafold has revealed novel interactions among attack phase-hypothetical proteins, which may be involved in less understood mechanisms of the Bdellovibrio attack phase. Here, we overlapped attack phase genes from B. bacteriovorus transcriptomic data sets and from transposon sequencing data sets to generate a set of proteins that are both expressed at the attack phase and are necessary for predation, which we termed Attack Phase Predation-Essential Proteins (AP-PEP). By applying Markov Cluster Algorithm and AlphaFold-Multimer to analyze the protein network and interaction partners of the AP-PEPs, we predicted high-confidence protein-protein interactions and two structurally similar but unique novel protein complexes formed among proteins of the Bd2209-Bd2212 and Bd2723-Bd2726 operons. Furthermore, we confirmed the interaction between hypothetical proteins Bd0075 and Bd0474 using the Bacteria Adenylate Cyclase Two-Hybrid system. In addition, we confirmed that the C-terminal domain of Bd0075, which contains Tetratricopeptide repeat motifs, participates principally in its interaction with Bd0474. This study revealed previously unknown cooperation among predation essential hypothetical proteins in the attack phase B. bacteriovorus and has paved the way for further work to understand molecular mechanisms of BALO predation processes.

Fitting Atomic Structures into Cryo-EM Maps by Coupling Deep Learning-Enhanced Map Processing with Global-Local Optimization

With the breakthroughs in protein structure prediction technology, constructing atomic structures from cryo-electron microscopy (cryo-EM) density maps through structural fitting has become increasingly critical. However, the accuracy of the constructed models heavily relies on the precision of the structure-to-map fitting. In this study, we introduce DEMO-EMfit, a progressive method that integrates deep learning-based backbone map extraction with a global-local structural pose search to fit atomic structures into density maps. DEMO-EMfit was extensively evaluated on a benchmark data set comprising both cryo-electron tomography (cryo-ET) and cryo-EM maps of protein and nucleic acid complexes. The results demonstrate that DEMO-EMfit outperforms state-of-the-art approaches, offering an efficient and accurate tool for fitting atomic structures into density maps.

Discovery of diverse and high-quality mRNA capping enzymes through a language model-enabled platform

Mining and expanding high-quality genetic parts for synthetic biology and bioengineering are urgent needs in the research and development of next-generation biotechnology. However, gene mining has relied on sequence homology or ample expert knowledge, which fundamentally limits the establishment of a comprehensive genetic part catalog. In this work, we propose SYMPLEX (synthetic biological part mining platform by large language model-enabled knowledge extraction), a universal gene-mining platform based on large language models. We applied SYMPLEX to mine enzymes responsible for messenger RNA (mRNA) capping, a key process in eukaryotic posttranscriptional modification, and obtained thousands of diverse candidates with traceable evidence from biomedical literature and databases. Of the 46 experimentally tested integral capping enzyme candidates, 14 demonstrated in vivo cross-species capping activity, and 2 displayed superior in vitro activity over the commercial vaccinia capping enzymes currently used in mRNA vaccine production. SYMPLEX provides a distinct paradigm for functional gene mining and offers powerful tools to facilitate knowledge discovery in fundamental research.

Conserved energetic changes drive function in an ancient protein fold

Many protein sequences occupy similar three-dimensional structures known as protein folds. In nature, protein folds are well-conserved over the course of evolution, such that there are 100,000 times as many extant protein sequences than there are folds. Despite their common shapes, similar protein folds can adopt wide-ranging functions, raising the question: are protein folds so strongly conserved for the purpose of maintaining function-driving energetic changes in protein families? Here we show that a set of key energetic relationships in a family of bacterial transcription factors (TFs) is conserved using high-resolution hydrogen exchange/mass spectrometry, bioinformatics, X-ray crystallography, and molecular dynamics simulations. We compared the TFs to their anciently diverged structural homologs, the periplasmic binding proteins (PBPs), expecting that protein families that share the same fold and bind the same sugars would have similar energetic responses. Surprisingly, our findings reveal the opposite: the "energetic blueprints" of the PBPs and the TFs are largely distinct, with the allosteric network of the TFs evolving specifically to support the functional requirements of genome regulation, versus conserved interactions with membrane transport machinery in PBPs. These results demonstrate how the same fold can be adapted for different sense/response functions via family-specific energetic requirements - even when responding to the same chemical trigger. Understanding the evolutionarily conserved energetic blueprint for a protein family provides a roadmap for designing functional proteins de novo , and will help us treat aberrant protein behavior in conserved domains in disease-related drug targets, where engineering selectivity is challenging.

Assessing interface accuracy in macromolecular complexes

Accurately predicting the 3D structures of macromolecular complexes is becoming increasingly important for understanding their cellular functions. At the same time, reliably assessing prediction quality remains a significant challenge in bioinformatics. To address this, various methods analyze and evaluate in silico models from multiple perspectives, accounting for both the reconstructed components' structures and their arrangement within the complex. In this work, we introduce Intermolecular Interaction Network Fidelity (I-INF), a normalized similarity measure that quantifies intermolecular interactions in multichain complexes. Adapted from a well-established score in the RNA field, I-INF provides a clear and intuitive way to evaluate the predicted 3D models against a reference structure, with a specific focus on interchain interaction sites. Additionally, we implement the F1 measure to assess interfaces in macromolecular assemblies, further enriching the evaluation framework. Tested on 72 RNA-protein decoys, as well as exemplary DNA-DNA, RNA-RNA, and protein-protein complexes, these measures deliver reliable scores and enable straightforward ranking of predictions. The tool for computing I-INF and F1 is publicly available on Zenodo, facilitating large-scale analysis and integration with other computational systems.

RNA sculpting by the primordial Helix-clasp-Helix-Strand-Loop (HcH-SL) motif enforces chemical recognition enabling diverse KH domain functions

In all domains of life, the ancient K homology (KH) domain superfamily is central to RNA processes including splicing, transcription, posttranscriptional gene regulation, signaling, and translation. Proteins with 1 to 15 KH domains bind single-strand (ss) RNA or DNA with base sequence specificity. Here, we examine over 40 KH domain experimental structures in complex with nucleic acid (NA) and define a novel Helix-clasp-Helix-Strand-Loop (HcH-SL) NA recognition motif binding 4 to 5 nucleotides using 10 to 18 residues. HcH-SL includes and extends the Gly-X-X-Gly (GXXG) signature sequence "clasp" that brings together two helices as an ∼90° helical corner. The first helix primarily provides side chain interactions to unstack and sculpt 2 to 3 bases on the 5' end for recognition of sequence and chemistry. The clasp and second helix amino dipole recognize a central phosphodiester. Following the helical corner, a beta strand and its loop extension recognize the two 3' nucleotides, primarily through main chain interactions. The HcH-SL structural motif forms a right-handed triangle and concave functional interface for NA interaction that unexpectedly splays four bound nucleotides into conformations matching RNA recognition motif (RRM) bound RNA structures. Evolutionary analyses and its ability to recognize base sequence and chemistry make HcH-SL a primordial NA binding motif distinguished by its binding mode from other NA structural recognition motifs: helix-turn-helix, helix-hairpin-helix, and beta strand RRM motifs. Combined results explain its vulnerability as a viral hijacking target and how mutations and expression defects lead to diverse diseases spanning cancer, cardiovascular, fragile X syndrome, neurodevelopmental disorders, and paraneoplastic disease.

Deciphering bacterial protein functions with innovative computational methods

Bacteria colonize every niche on Earth and play key roles in many environmental and host-associated processes. The sequencing revolution revealed the remarkable bacterial genetic and proteomic diversity and the genomic content of cultured and uncultured bacteria. However, deciphering functions of novel proteins remains a high barrier, often preventing the deep understanding of microbial life and its interaction with the surrounding environment. In recent years, exciting new bioinformatic tools, many of which are based on machine learning, facilitate the challenging task of gene and protein function discovery in the era of big genomics data, leading to the generation of testable hypotheses for bacterial protein functions. The new tools allow prediction of protein structures and interactions and allow sensitive and efficient sequence- and structure-based searching and clustering. Here, we summarize some of these recent tools which revolutionize modern microbiology research, along with examples for their usage, emphasizing the user-friendly, web-based ones. Adoption of these capabilities by experimentalists and computational biologists could save resources and accelerate microbiology research.

A Sorghum BAK1/SERK4 Homolog Functions in Pathogen-Associated Molecular Patterns-Triggered Immunity and Cell Death in Response to Colletotrichum sublineola Infection

Sorghum bicolor is the fifth most important cereal crop and expected to gain prominence due to its versatility, low input requirements, and tolerance to hot and dry conditions. In warm and humid environments, the productivity of sorghum is severely limited by the hemibiotrophic fungal pathogen Colletotrichum sublineola, the causal agent of anthracnose. Cultivating anthracnose-resistant accessions is the most effective and environmentally benign way to safeguard yield. A previous genome-wide association study for anthracnose resistance in the Sorghum Association Panel uncovered single-nucleotide polymorphisms on chromosome 5 associated with resistance to anthracnose, including one located within the coding region of gene Sobic.005G182400. In this study, we investigated the molecular function of Sobic.005G182400 in response to C. sublineola infection. Conserved domain, phylogenetic, and structural analyses revealed that the protein encoded by Sobic.005G182400 shares significant structural similarity with the Arabidopsis BRASSINOSTEROID INSENSITIVE1-ASSOCIATED RECEPTOR KINASE1 (BAK1)/SOMATIC EMBRYOGENESIS RECEPTOR-LIKE KINASE4 (SERK4). Although sequence analysis of four sorghum accessions showed no substantial variation in the coding region, accession SC1330, which carries the resistance allele, exhibited significantly higher expression of Sobic.005G182400 during early infection (≤24 h). Co-expression network analysis identified that the module associated with Sobic.005G182400 was enriched in genes involved in endocytosis, autophagy, and vesicle transport. Gene regulatory network analysis further suggested that Sobic.005G182400 regulates genes required for BAK1/SERK4-mediated cell death via protein glycosylation. Together, these findings indicate that Sobic.005G182400 encodes a protein with similarity to Arabidopsis BAK1/SERK4 that enables pathogen-associated molecular patterns-triggered immunity and regulates cell death.

Evolutionary landscape of plant chalcone isomerase-fold gene families

Flavonoids are crucial for plant survival and adaptive evolution, and chalcone isomerase (CHI) genes serve as key rate-limiting gene in the flavonoid biosynthesis pathway. It is important for plant adaptive evolution to comprehensively study the evolution and diversity of the CHI gene families. However, the CHI gene families in many plant lineages remain elusive. This study systematically identified CHI genes from 259 species including algae, bryophytes, ferns, gymnosperms, and angiosperms. A total of 1,738 CHI gene family members were discovered. We analyzed the diversity, distribution trajectory, and the driving forces of gene duplication during the evolution of the plant lineages. The present study is the first to identify potential type II and type IV CHI genes in the extant liverwort model species Marchantia polymorpha. The distribution pattern of CHI genes across the plant kingdom reveals that the origin of type II CHI can be traced back to the last common ancestor of bryophytes and vascular plants, and type III CHI may represent the ancestral form of the CHI gene family. The identification of conserved motifs showed significant differences in motif distribution among different CHI gene types. It was found that the drivers of gene duplication varied across plant lineages: dispersed duplications (DSD) were predominant in algae and bryophytes, whole-genome duplication (WGD) was the main driver in basal angiosperms and monocots, while tandem duplications (TD) predominating in eudicots. Structural clustering analysis demonstrated the 3-layer sandwich structure in the CHI-fold proteins remained conserved in the central region, while repeated loss of N-terminal sequences contributed to structural diversity. This study provides a deeper understanding of the evolution and diversity of the CHI-fold proteins and lays a theoretical foundation for further studies of their function and the identification of new functional CHI genes.

Mining Highly Active Oleate Hydratases by Structure Clustering, Sequence Clustering, and Ancestral Sequence Reconstruction

Oleate hydratases (Ohys) catalyze the conversion of oleic acid (OA) to 10-(R)-hydroxystearic acid (10-HSA), a compound widely used in the chemical industry. However, the limited activity of Ohys has hindered their broader applications. To address this limitation, we propose a novel strategy for mining highly active Ohys through structure clustering, sequence clustering, and ancestral sequence reconstruction (SSA strategy). Structure clustering via AI-driven protein structure prediction followed by classification enhanced the ability to mine target Ohys. Ancestral enzyme reconstruction was carried out based on mining results from both structure and sequence clustering. This strategy significantly reduces the time and cost of the discovery process. Among the 1304 Ohys screened via SSA, 13 candidates were selected. Seven candidates demonstrated high activity. Ohy 64, identified through structure clustering, exhibited the highest activity. Ancestral enzymes that were reconstructed from structure clustering targets were 3 times more likely to exhibit high catalytic activity than those identified through sequence clustering. Four critical, hydrophobic residues were identified through structure and sequence comparisons between StOhy and targets mined by SSA. Site-directed mutagenesis revealed that these hydrophobic residues conferred varying levels of enzyme activity, confirming that increased hydrophobicity at these positions enhances cofactor FAD binding, thus improving enzyme catalytic efficiency.

Rational Introduction of Electrostatic Interactions at Crystal Contacts to Enhance Protein Crystallization of an Ene Reductase

Protein crystallization is an alternative to well-established but cost-intensive and time-consuming chromatography in biotechnological processes, with protein crystallization defined as an essential unit operation for isolating proteins, e.g., active pharmaceutical ingredients. Crystalline therapeutic proteins attract interest in formulation and delivery processes of biopharmaceuticals due to the high purity, concentration, and stability of the crystalline state. Although improving protein crystallization is mainly achieved by high-throughput screening of crystallization conditions, recent studies have established a rational protein engineering approach to enhance crystallization for two homologous alcohol dehydrogenases from Lactobacillus brevis (LbADH) and Lactobacillus kefiri (LkADH). As generalizing crystallization processes across a wide range of target proteins remains challenging, this study takes a further step by applying the successful crystal contact engineering strategies for LbADH/LkADH to a non-homologous protein, an NADH-binding derivative of the Nostoc sp. PCC 1720 ene reductase (NspER1-L1,5). Here, the focus lies on introducing electrostatic interactions at crystal contacts, specifically between lysine and glutamic acid. Out of the nine tested NspER1-L1,5 mutants produced in E. coli, six crystallized, while four mutants revealed an increased propensity to crystallize in static µL-batch crystallization compared to the wild type: Q204K, Q350K, D352K, and T354K. The best-performing mutant Q204K was selected for upscaling, crystallizing faster than the wild type in a stirred batch crystallizer. Even when spiked with E. coli cell lysate, the mutant maintained increased crystallizability compared to the wild type. The results of this study highlight the potential of crystal contact engineering as a reliable tool for improving protein crystallization as an alternative to chromatography, paving the way for more efficient biotechnological downstream processing.

Structural-functional analyses of the huntingtin/HAP40 complex in Drosophila and humans

Huntington's disease (HD) is a neurodegenerative disorder caused by an abnormal CAG expansion in the Huntingtin (HTT) gene. Given its simple genetic cause but complex pathogenic mechanisms, interest in targeting HTT for HD treatment is growing, necessitating a clear understanding of HTT regulation. HTT protein primarily exists in a core complex with HAP40, forming a highly ordered structure with two large globular domains connected by a bridge. We previously demonstrated that HAP40 is conserved in Drosophila, controls HTT's function, protein stability, and levels, and is a potential modifier of HD pathogenesis, supporting its central role in HTT regulation. Here, we showed that HTT synergizes with HAP40 to induce novel gain-of-function effects in Drosophila when overexpressed. Protein modeling revealed that despite their prominent evolutionary and sequence divergence, the fly and human HTT-HAP40 complexes share a high degree of structural similarity. Protein-contact maps and molecular simulations showed that HAP40 preferentially binds to HTT's C-terminal domain in both complexes. By examining the interfacial contacts between HTT and HAP40 in fly and human complexes, we identified ten conserved bonds that are important for HAP40's affinity for HTT. Finally, we showed that the conserved N-terminal BΦ motif in HAP40 is not essential for HTT binding but important for HAP40's functions. Through the structural-functional analyses of the fly and human HTT-HAP40 complexes, our results support that the structural similarity underlies the functional conservation of the two complexes from these evolutionarily distant species and further uncover novel insight into HAP40 regulation and its interaction with HTT.

Multi-scale structural similarity embedding search across entire proteomes

The rapid expansion of three-dimensional (3D) biomolecular structure information, driven by breakthroughs in artificial intelligence/deep learning (AI/DL)-based structure predictions, has created an urgent need for scalable and efficient structure similarity search methods. Traditional alignment-based approaches, such as structural superposition tools, are computationally expensive and challenging to scale with the vast number of available macromolecular structures. Herein, we present a scalable structure similarity search strategy designed to navigate extensive repositories of experimentally determined structures and computed structure models predicted using AI/DL methods. Our approach leverages protein language models and a deep neural network architecture to transform 3D structures into fixed-length vectors, enabling efficient large-scale comparisons. Although trained to predict TM-scores between single-domain structures, our model generalizes beyond the domain level, accurately identifying 3D similarity for full-length polypeptide chains and multimeric assemblies. By integrating vector databases, our method facilitates efficient large-scale structure retrieval, addressing the growing challenges posed by the expanding volume of 3D biostructure information.

Two Methods for Superposing the Structures of Like-Molecule Assemblies: Application to Peptide and Protein Oligomers and Aggregates

Two algorithms are proposed for the superposition of assemblies of like molecules (e.g., peptide and proteins homooligomers and homoaggregates), which do not require examining all permutations of the molecules. Both start from searching the mutual orientation of the two assemblies over a grid of quaternion components for the sub-optimal mapping and orientation of the molecules of the second to those of the first assembly. The first one, termed Like-Molecule Assembly Distance Alignment (LMADA), uses Singular Value Decomposition to superpose the two assemblies, given the sub-optimal mapping. The second one, termed Like-Molecule Assembly Gaussian Distance Alignment (LMAGDA), minimizes the negative of the logarithm of the sum of the Gaussian terms in the distances between the corresponding atoms/sites of all pairs of molecules of the two assemblies in quaternion components, starting from those estimated in the first stage. Both algorithms yield as good or nearly as good superposition, in terms of root mean square deviation (RMSD), as examining all permutations to find the lowest RMSD. LMADA results in lower RMSDs, while LMAGDA in a better alignment of the geometrically matching sections of the assemblies. The costs of the proposed algorithms scale only with N2, N being the number of molecules in the assembly, as opposed to N! when examining all permutations.

Rapid and sensitive protein complex alignment with Foldseek-Multimer

Advances in computational structure prediction will vastly augment the hundreds of thousands of currently available protein complex structures. Translating these into discoveries requires aligning them, which is computationally prohibitive. Foldseek-Multimer computes complex alignments from compatible chain-to-chain alignments, identified by efficiently clustering their superposition vectors. Foldseek-Multimer is 3-4 orders of magnitudes faster than the gold standard, while producing comparable alignments; this allows it to compare billions of complex pairs in 11 h. Foldseek-Multimer is open-source software available at GitHub via https://github.com/steineggerlab/foldseek/ , https://search.foldseek.com/search/ and the BFMD database.

Aging-induced Alternation in the Gut Microbiota Impairs Host Antibacterial Defense

Older individuals experience increased susceptibility and mortality to bacterial infections, but the underlying etiology remains unclear. Herein, it is shown that aging-associated reduction of commensal Parabacteroides goldsteinii (P. goldsteinii) in both aged mice and humans critically contributes to worse outcomes of bacterial infection. The colonization of live P. goldsteinii conferred protection against aging-associated bacterial infections. Metabolomic profiling reveals a protective compound, apigenin, generated by P. goldsteinii, antagonizes bacterial clearance defects in aged mice. AMP-binding protein (ampB) is identified as a key gene involved in apigenin synthesis in P. goldsteinii using homologous recombination in bacteria. Mechanistically, apigenin binds directly to the potential sites on Fgr (M341 and D404), preventing its inhibitory role on Vav1 phosphorylation, and therefore promoting the activation of Cdc42/Rac1, Arp2/3 expression and subsequent actin reorganization, which contributes to the enhanced phagocytosis of macrophages to bacteria. Collectively, the findings suggest that dysbiosis of the gut microbiota may impair host defense mechanisms and increase susceptibility to bacterial infections in older adults and highlight the microbiota-apigenin-Fgr axis as a possible route to ameliorate aging-associated antibacterial defects.

Computational studies reveal structural characterization and novel families of Puccinia striiformis f. sp. tritici effectors

Understanding the biological functions of Puccinia striiformis f. sp. tritici (Pst) effectors is fundamental for uncovering the mechanisms of pathogenicity and variability, thereby paving the way for developing durable and effective control strategies for stripe rust. However, due to the lack of an efficient genetic transformation system in Pst, progress in effector function studies has been slow. Here, we modeled the structures of 15,201 effectors from twelve Pst races or isolates, a Puccinia striiformis isolate, and one Puccinia striiformis f. sp. hordei isolate using AlphaFold2. Of these, 8,102 folds were successfully predicted, and we performed sequence- and structure-based annotations of these effectors. These effectors were classified into 410 structure clusters and 1,005 sequence clusters. Sequence lengths varied widely, with a concentration between 101-250 amino acids, and motif analysis revealed that 47% and 5.81% of the predicted effectors contain known effector motifs [Y/F/W]xC and RxLR, respectively highlighting the structural conservation across a substantial portion of the effectors. Subcellular localization predictions indicated a predominant cytoplasmic localization, with notable chloroplast and nuclear presence. Structure-guided analysis significantly enhances effector prediction efficiency as demonstrated by the 75% among 8,102 have structural annotation. The clustering and annotation prediction both based on the sequence and structure homologies allowed us to determine the adopted folding or fold families of the effectors. A common feature observed was the formation of structural homologies from different sequences. In our study, one of the comparative structural analyses revealed a new structure family with a core structure of four helices, including Pst27791, PstGSRE4, and PstSIE1, which target key wheat immune pathway proteins, impacting the host immune functions. Further comparative structural analysis showed similarities between Pst effectors and effectors from other pathogens, such as AvrSr35, AvrSr50, Zt-KP4-1, and MoHrip2, highlighting a possibility of convergent evolutionary strategies, yet to be supported by further data encompassing on some evolutionarily distant species. Currently, our initial analysis is the most one on Pst effectors' sequence, structural and annotation relationships providing a novel foundation to advance our future understanding of Pst pathogenicity and evolution.

Comparative Analyses of Compatible and Incompatible Host-Pathogen Interactions Provide Insight into Divergent Host Specialization of Closely Related Pathogens

Host-pathogen co-evolutionary dynamics drive constant changes in plant pathogens to thrive in their plant host. Factors that determine host specificity are diverse and range from molecular and morphological strategies to metabolic and reproductive adaptations. We applied an experimental approach and conducted comparative microscopy, transcriptome analyses, and functional analyses of selected pathogen traits to identify determinants of host specificity in an important wheat pathogen. We included three closely related fungal pathogens, Zymoseptoria tritici, Z. pseudotritici, and Z. ardabiliae, that establish compatible and incompatible interactions with wheat. Although infections of the incompatible species induce plant defenses during invasion of stomatal openings, we found a conserved early-infection program among the three species whereby only 9.2% of the 8,885 orthologous genes are significantly differentially expressed during initial infection. The genes upregulated in Z. tritici likely reflect specialization to wheat, whereas upregulated genes in the incompatible interaction may reflect processes to counteract cellular stress associated with plant defenses. We selected nine candidate genes encoding putative effectors and host-specificity determinants in Z. tritici and deleted these to study their functional relevance. Despite the particular expression patterns of the nine genes, only two mutants were impaired in virulence. We further expressed the Z. tritici proteins in Nicotiana benthamiana to investigate protein function and assess cell death reaction. Hereby, we identify three effectors with cell-death-inducing properties. From the functional analyses, we conclude that the successful infection of Z. tritici in wheat relies on an extensive redundancy of virulence determinants. [Formula: see text] Copyright © 2025 The Author(s). This is an open access article distributed under the CC BY-NC-ND 4.0 International license.

Secondary-Structure-Informed RNA Inverse Design via Relational Graph Neural Networks

RNA inverse design is an essential part of many RNA therapeutic strategies. To date, there have been great advances in computationally driven RNA design. The current machine learning approaches can predict the sequence of an RNA given its 3D structure with acceptable accuracy and at tremendous speed. The design and engineering of RNA regulators such as riboswitches, however, is often more difficult, partly due to their inherent conformational switching abilities. Although recent state-of-the-art models do incorporate information about the multiple structures that a sequence can fold into, there is great room for improvement in modeling structural switching. In this work, a relational geometric graph neural network is proposed that explicitly incorporates alternative structures to predict an RNA sequence. Converting the RNA structure into a geometric graph, the proposed model uses edge types to distinguish between the primary structure, secondary structure, and spatial positioning of the nucleotides in representing structures. The results show higher native sequence recovery rates over those of gRNAde across different test sets (eg. 72% vs. 66%) and a benchmark from the literature (60% vs. 57%). Secondary-structure edge types had a more significant impact on the sequence recovery than the spatial edge types as defined in this work. Overall, these results suggest the need for more complex and case-specific characterization of RNA for successful inverse design.

Estimating Protein Complex Model Accuracy Using Graph Transformers and Pairwise Similarity Graphs

Motivation

Estimation of protein complex structure accuracy is an essential step in protein complex structure prediction and is also important for users to select good structural models for various applications, such as protein function analysis and drug design. Despite the success of structure prediction methods such as AlphaFold2 and AlphaFold3, predicting the quality of predicted complex structures (structural models) and selecting top ones from large model pools remains challenging.

Results

We present GATE, a novel method that uses graph transformers on pairwise model similarity graphs to predict the quality (accuracy) of complex structural models. By integrating single-model and multi-model quality features, GATE captures both the characteristics of individual models and the geometric similarity between them to make robust predictions. On the dataset of the 15th Critical Assessment of Protein Structure Prediction (CASP15), GATE achieved the highest Pearson's correlation (0.748) and the lowest ranking loss (0.1191) compared to existing methods. In the blind CASP16 experiment, GATE was ranked 4th according to the overall sum of z-scores of multiple metrics based on both TM-score and Oligo-GDTTS scores. In terms of per-target average metrics based on TM-score, GATE achieved a Pearson's correlation of 0.7076 (1st place among all methods), a Spearman's correlation of 0.4514 (3rd place), a ranking loss of 0.1221 (3rd place), and an Area Under the Curve (AUC) score of 0.6680 (3rd place), highlighting its strong, balanced ability of estimating complex model accuracy and selecting good models.

Availability

The source code of GATE is freely available at https://github.com/BioinfoMachineLearning/GATE.

RNA-protein interaction prediction using network-guided deep learning

Accurate computational determination of RNA-protein interactions remains challenging, particularly when encountering unknown RNAs and proteins. The limited number of RNAs and their flexibility constrained the effectiveness of the deep-learning models for RNA-protein interaction prediction. Here, we introduce ZHMolGraph, which integrates graph neural network and unsupervised large language models to predict RNA-protein interaction. We validate ZHMolGraph predictions on two benchmark datasets and outperform the current best methods. For the dataset of entirely unknown RNAs and proteins, ZHMolGraph shows an improvement in achieving high AUROC of 79.8% and AUPRC of 82.0%. This represents a substantial improvement of 7.1%-28.7% in AUROC and 4.6%-30.0% in AUPRC over other methods. We utilize ZHMolGraph to enhance the challenging SARS-CoV-2 RPI and unbound RNA-protein complex predictions. Such enhancements make ZHMolGraph a reliable option for genome-wide RNA-protein prediction. ZHMolGraph holds broad potential for modeling and designing RNA-protein complexes.

Functional and structural insights into α-L-Rhamnosidase: cloning, characterization, and decoding evolutionary constraints through structural motif

α-L-rhamnosidase [E.C. 3.2.1.40] is important in various industrial and biotechnological applications. However, limited knowledge of the structural features of its active site residues and their local geometric arrangements during substrate interaction hinders further application development. In this study, we examined functionally characterized microbial α-L-rhamnosidases. Despite considerable differences in their global structures, the local structures of the substrate-binding sites and key residues were highly conserved. Using the local structural motif, we characterized α-L-rhamnosidase genes from metagenomic samples of traditional fermentation starters. To comprehensively understand the distribution of α-L-rhamnosidases with this motif in the AlphaFold database, we screened 26,858 α-L-rhamnosidase structures. Our findings showed that only 5678 out of 26,858 structures contain the specific conserved motifs, emphasizing their potential significance in mining enzyme function. Moreover, the analysis of structural diversity among representative enzymes demonstrated variation in the number and types of domains within this enzyme family. Further investigation of representative α-L-rhamnosidase sequences with this structural motif confirmed the evolutionary constraints of 15 key residues, indicating strong selective pressures to maintain these elements essential for enzyme functionality. These residues were consistently present across ancestral sequences, underscoring their importance throughout the enzyme's evolutionary history. This study suggests that structure-guided approaches are valuable for discovering functional enzymes. Identifying conserved motif across diverse microbial taxa not only aids in predicting enzyme functionality but also offers opportunities for enzyme engineering and biotechnological applications.

Uncovering the hidden RNA virus diversity in Lake Nam Co: Evolutionary insights from an extreme high-altitude environment

Alpine lakes, characterized by isolation, low temperatures, oligotrophic conditions, and intense ultraviolet radiation, remain a poorly explored ecosystem for RNA viruses. Here, we present the first comprehensive metatranscriptomic study of RNA viruses in Lake Nam Co, a high-altitude alkaline saline lake on the Tibetan Plateau. Using a combination of sequence- and structure-based homology searches, we identified 742 RNA virus species, including 383 novel genus-level groups and 84 novel family-level groups exclusively found in Lake Nam Co. These findings significantly expand the known diversity of the Orthornavirae, uncovering evolutionary adaptations such as permutated RNA-dependent RNA polymerase motifs and distinct RNA secondary structures. Notably, 14 additional RNA virus families potentially infecting prokaryotes were predicted, broadening the known host range of RNA viruses and questioning the traditional assumption that RNA viruses predominantly target eukaryotes. The presence of auxiliary metabolic genes in viral genomes suggested that RNA viruses (families f.0102 and Nam-Co_family_51) exploit host energy production mechanisms in energy-limited alpine lakes. Low nucleotide diversity, single nucleotide polymorphism frequencies, and pN/pS ratios indicate strong purifying selection in Nam Co viral populations. Our findings offer insights into RNA virus evolution and ecology, highlighting the importance of extreme environments in uncovering hidden viral diversity and further shed light into their potential ecological implications, particularly in the context of climate change.

Accurate Prediction of Protein Complex Stoichiometry by Integrating AlphaFold3 and Template Information

Protein structure prediction methods require stoichiometry information (i.e., subunit counts) to predict the quaternary structure of protein complexes. However, this information is often unavailable, making stoichiometry prediction crucial for complexes with unknown stoichiometry. Despite its importance, few computational methods address this challenge. In this study, we present an approach that integrates AlphaFold3 structure predictions with homologous template data to predict stoichiometry. The method generates candidate stoichiometries, builds structural models for them using AlphaFold3, ranks them based on AlphaFold3 scores, and further refine predictions with template-based information when available. In the 16th community-wide Critical Assessment of Techniques for Protein Structure Prediction (CASP16), our method achieved 71.4% top-1 accuracy and 92.9% top-3 accuracy, outperforming other predictors in terms of the overall performance. This demonstrates the complementary strengths of AlphaFold3- and template-based predictions and highlights its applicability for uncharacterized protein complexes lacking stoichiometry data.

Faithful Interpretation of Protein Structures through Weighted Persistent Homology Improves Evolutionary Distance Estimation

Phylogenetic inference is mainly based on sequence analysis and requires reliable alignments. This can be challenging, especially when sequences are highly divergent. In this context, the use of three-dimensional protein structures is a promising alternative. In a recent study, we introduced an original topological data analysis method based on persistent homology to estimate the evolutionary distances from structures. The method was successfully tested on 518 protein families representing 22,940 predicted structures. However, as anticipated, the reliability of the estimated evolutionary distances was impacted by the quality of the predicted structures and the presence of indels in the proteins. This paper introduces a new topological descriptor, called bio-topological marker (BTM), which provides a more faithful description of the structures, a topological analysis for estimating evolutionary distances from BTMs, and a new weight-filtering method adapted to protein structures. These new developments significantly improve the estimation of evolutionary distances and phylogenies inferred from structures.

Tandem Repeats Provide Evidence for Convergent Evolution to Similar Protein Structures

Homology is a key concept underpinning the comparison of sequences across organisms. Sequence-level homology is based on a statistical framework optimized over decades of work. Recently, computational protein structure prediction has enabled large-scale homology inference beyond the limits of accurate sequence alignment. In this regime, it is possible to observe nearly identical protein structures lacking detectable sequence similarity. In the absence of a robust statistical framework for structure comparison, it is largely assumed similar structures are homologous. However, it is conceivable that matching structures could arise through convergent evolution, resulting in analogous proteins without shared ancestry. Large databases of predicted structures offer a means of determining whether analogs are present among structure matches. Here, I find that a small subset (∼2.6%) of Foldseek clusters lack sequence-level support for homology, including ∼1% of strong structure matches with template modeling score ≥ 0.5. This result by itself does not imply these structure pairs are nonhomologous, since their sequences could have diverged beyond the limits of recognition. Yet, strong matches without sequence-level support for homology are enriched in structures with predicted repeats that could induce spurious matches. Some of these structural repeats are underpinned by sequence-level tandem repeats in both matching structures. I show that many of these tandem repeat units have genealogies inconsistent with their corresponding structures sharing a common ancestor, implying these highly similar structure pairs are analogous rather than homologous. This result suggests caution is warranted when inferring homology from structural resemblance alone in the absence of sequence-level support for homology.

I / σ I vs {Rmerg, Rmeas, Rpim, CC1/2} for Crystal Diffraction Data Quality Evaluation

X-ray crystal diffraction has provided atomic-level structural information on biological macromolecules. Data quality determines the reliability of structural models. In most cases, multiple data sets are available from different crystals and/or collected with different experimental settings. Reliable metrics are critical to rank and select the data set with the highest quality. Many measures have been created or modified for data quality evaluation. However, some are duplicate in functionality, and some are likely misused due to misunderstanding, which causes confusion or problems, especially at synchrotron beamlines where experiments proceed quickly. In this work, these measures are studied through both theoretical analysis and experimental data with various characteristics, which demonstrated that: 1). {Rmerg, Rmeas, Rpim, CC1/2} all measure the equivalence of reflections, and the low-shell values of these metrics can be used as reliable indicators for correctness (or trueness) of Laue symmetry; 2). High-shell I / σ I is a reliable and better indicator to select resolution cutoff while the overall value measures the overall strength of the data.

Oxidative adaptations in prokaryotes imply the oxygenic photosynthesis before crown-group Cyanobacteria

The metabolic transition from anaerobic to aerobic in prokaryotes reflects adaptations to oxidative stress. Methanogen, one of the earliest life forms on Earth, has evolved into three major groups within the Euryarchaeota, exhibiting different phylogenetic affiliations and metabolic characters. In comparison with other strictly anaerobic methanogenic groups, the Class II methanogens possess a better capability to adapt to limited oxygen pressure. Cyanobacteria is considered the first and only prokaryote evolving oxygenic photosynthesis and is responsible for the Great Oxidation Event on Earth. However, the connection between oxygenic Cyanobacteria and evolutionary adaptations to oxidative stress in prokaryotes remains elusive. Here, through the gene encoding structural maintenance of chromosomes (SMC) protein, which was horizontally transferred from ancient Class II methanogens to the last common ancestor of the crown-group Cyanobacteria, we demonstrate that the origin of extant Cyanobacteria was undoubtedly posterior to the occurrence of oxygen-tolerant Class II methanogens. In addition, we found that certain prokaryotic lineages had evolved the tolerance mechanisms against oxidative stress before the origin of extant Cyanobacteria. The contradiction that oxidative adaptations in Class II methanogens and other prokaryotes predating the crown-group oxygenic Cyanobacteria implies the existence of more ancient biological oxygenesis. We propose that these potential oxygenic organisms might represent the extinct phototrophs and first emerge during the Paleoarchean, contributing to the oxidative adaptations in the prokaryotic tree of life and facilitating the dispersal of reaction centers across the bacterial domain.

Evidence for corrin biosynthesis in the last universal common ancestor

Corrinoids are cobalt-containing tetrapyrroles. They include adenosylcobalamin (vitamin B12) and cobamides that function as cofactors and coenzymes for methyl transfer, radical-dependent and redox reactions. Though cobamides are the most complex cofactors in nature, they are essential in the acetyl-CoA pathway, thought to be the most ancient CO2-fixation pathway, where they perform a pterin-to-cobalt-to-nickel methyl transfer reaction catalyzed by the corrinoid iron-sulphur protein (CoFeS). CoFeS occurs in H2-dependent archaeal methanogens, the oldest microbial lineage by measure of physiology and carbon isotope data, dating corrinoids to ca. 3.5 billion years. However, CoFeS and cobamides are also essential in the acetyl-CoA pathway of H2-dependent bacterial acetogens. To determine whether corrin biosynthesis was established before archaea and bacteria diverged, whether the pathways arose independently or whether cobamide biosynthesis was transferred from the archaeal to the bacterial lineage (or vice versa) during evolution, we investigated phylogenies and structural data for 26 enzymes of corrin ring and lower ligand biosynthesis. The data trace cobamide synthesis to the common ancestor of bacteria and archaea, placing it in the last universal common ancestor of all lifeforms (LUCA), while pterin-dependent methyl synthesis pathways likely arose independently post-LUCA in the lineages leading to bacteria and archaea. Enzymes of corrin biosynthesis were recruited from preexisting ancient pathways. Evolutionary forerunners of CoFeS function were likely Fe-, Ni- and Co-containing solid-state surfaces, which, in the laboratory, catalyze the reactions of the acetyl-CoA pathway from CO2 to pyruvate under serpentinizing hydrothermal conditions. The data suggest that enzymatic corrin biosynthesis replaced insoluble solid-state catalysts that tethered primordial CO2 assimilation to the Earth's crust, suggesting a role for corrin synthesis in the origin of free-living cells.

Increasing thermostability of the key photorespiratory enzyme glycerate 3-kinase by structure-based recombination

As global temperatures rise, improving crop yields will require enhancing the thermotolerance of crops. One approach for improving thermotolerance is using bioengineering to increase the thermostability of enzymes catalysing essential biological processes. Photorespiration is an essential recycling process in plants that is integral to photosynthesis and crop growth. The enzymes of photorespiration are targets for enhancing plant thermotolerance as this pathway limits carbon fixation at elevated temperatures. We explored the effects of temperature on the activity of the photorespiratory enzyme glycerate kinase (GLYK) from various organisms and the homologue from the thermophilic alga Cyanidioschyzon merolae was more thermotolerant than those from mesophilic plants, including Arabidopsis thaliana. To understand enzyme features underlying the thermotolerance of C. merolae GLYK (CmGLYK), we performed molecular dynamics simulations using AlphaFold-predicted structures, which revealed greater movement of loop regions of mesophilic plant GLYKs at higher temperatures compared to CmGLYK. Based on these simulations, hybrid proteins were produced and analysed. These hybrid enzymes contained loop regions from CmGLYK replacing the most mobile corresponding loops of AtGLYK. Two of these hybrid enzymes had enhanced thermostability, with melting temperatures increased by 6 °C. One hybrid with three grafted loops maintained higher activity at elevated temperatures. Whilst this hybrid enzyme exhibited enhanced thermostability and a similar Km for ATP compared to AtGLYK, its Km for glycerate increased threefold. This study demonstrates that molecular dynamics simulation-guided structure-based recombination offers a promising strategy for enhancing the thermostability of other plant enzymes with possible application to increasing the thermotolerance of plants under warming climates.

Has AlphaFold3 achieved success for RNA?

Predicting the 3D structure of RNA is a significant challenge despite ongoing advancements in the field. Although AlphaFold has successfully addressed this problem for proteins, RNA structure prediction raises difficulties due to the fundamental differences between proteins and RNA, which hinder its direct adaptation. The latest release of AlphaFold, AlphaFold3, has broadened its scope to include multiple different molecules such as DNA, ligands and RNA. While the AlphaFold3 article discussed the results for the last CASP-RNA data set, the scope of its performance and the limitations for RNA are unclear. In this article, we provide a comprehensive analysis of the performance of AlphaFold3 in the prediction of 3D structures of RNA. Through an extensive benchmark over five different test sets, we discuss the performance and limitations of AlphaFold3. We also compare its performance with ten existing state-of-the-art ab initio, template-based and deep-learning approaches. Our results are freely available on the EvryRNA platform at https://evryrna.ibisc.univ-evry.fr/evryrna/alphafold3/.

PSSKB: A Web Application to Study Protein Structures

The proteins expressed during the cell cycle determine cell function and ensure signaling pathway activation in response to environmental influences. Developments in structural biology, biophysics, and bioinformatics provide information on the structure and function of particular proteins including that on the structural changes in proteins due to post-translational modification (PTM) and amino acid substitutions (AAS), which is essential for understanding protein function and life cycle. These are PTMs and AASs that often modulate the function and alter the stability and localization of a protein in a cell. PSSKB is a platform that integrates all necessary tools for modeling the five common natural modifications and all canonical AASs in proteins. The available tools are not limited to the local database, so the user can select a protein from Uniprot ID or PDB ID. The result will be a three-dimensional (3D) representation of the modified structure, as well as an analysis of the changes in the performance of the intact and modified structures after energy minimization compared with the original structure, which not only makes it possible to evaluate AAS/PTM influence of on a protein's characteristics but also to use the 3D model for further studies. Additionally, PSSKB enables the user to search, align, overlay, and determine the exact coordinates of protein structure fragments. The search results are a set of structural motifs similar to the query and ranked by statistical significance. The platform is fully functional and publicly available at https://psskb.org/. No registration is required to access the platform. A tutorial video can be found at https://psskb.org/page/about. Services provided on the platform are based on previously developed and published software. SCPacker applied for PTM Modeling and AAS services available at GitHub (https://github.com/protdb/SCPacker). SaFoldNet applied for a Similar Search service is also available at GitHub (https://github.com/protdb/ABBNet).

A comprehensive survey of scoring functions for protein docking models

BACKGROUND

While protein-protein docking is fundamental to our understanding of how proteins interact, scoring protein-protein complex conformations is a critical component of successful docking programs. Without accurate and efficient scoring functions to differentiate between native and non-native binding complexes, the accuracy of current docking tools cannot be guaranteed. Although many innovative scoring functions have been proposed, a good scoring function for docking remains elusive. Deep learning models offer alternatives to using explicit empirical or mathematical functions for scoring protein-protein complexes.

RESULTS

In this study, we perform a comprehensive survey of the state-of-the-art scoring functions by considering the most popular and highly performant approaches, both classical and deep learning-based, for scoring protein-protein complexes. The methods were also compared based on their runtime as it directly impacts their use in large-scale docking applications.

CONCLUSIONS

We evaluate the strengths and weaknesses of classical and deep learning-based approaches across seven public and popular datasets to aid researchers in understanding the progress made in this field.

Single-cell transcriptome atlas of lamprey exploring Natterin- induced white adipose tissue browning

Lampreys are early jawless vertebrates that are the key to understanding the evolution of vertebrates. However, the lack of cytomic studies on multiple lamprey organs has hindered progress in this field. Therefore, the present study constructed a comprehensive cell atlas comprising 604,460 cells/nuclei and 70 cell types from 14 lamprey tissue samples. Comparison of cellular evolution across species revealed that most lamprey cell types are homologous to those in jawed vertebrates. We discovered acinar- and islet-like cell populations despite the lack of parenchymal organs in lampreys, providing evidence of pancreatic function in vertebrates. Furthermore, we investigated the heterogeneity of lamprey immune cell populations. Natterin was highly expressed in granulocytes, and NATTERIN was localized to the lipid droplets. Moreover, we developed a transgenic mouse model expressing Natterin to elucidate the role of NATTERIN in lipid metabolism, whereas the browning of white adipose tissue was induced. These findings elucidate vertebrate cellular evolution and advance our understanding of adipose tissue plasticity and metabolic regulation in lampreys.

Accurate Prediction of Protein Complex Stoichiometry by Integrating AlphaFold3 and Template Information

Protein structure prediction methods require stoichiometry information (i.e., subunit counts) to predict the quaternary structure of protein complexes. However, this information is often unavailable, making stoichiometry prediction crucial for complexes with unknown stoichiometry. Despite its importance, few computational methods address this challenge. In this study, we present an approach that integrates AlphaFold3 structure predictions with homologous template data to predict stoichiometry. The method generates candidate stoichiometries, builds structural models for them using AlphaFold3, ranks them based on AlphaFold3 scores, and further refine predictions with template-based information when available. In the 16th community-wide Critical Assessment of Techniques for Protein Structure Prediction (CASP16), our method achieved 71.4% top-1 accuracy and 92.9% top-3 accuracy, outperforming other predictors in terms of the overall performance. This demonstrates the complementary strengths of AlphaFold3- and template-based predictions and highlights its applicability for uncharacterized protein complexes lacking stoichiometry data.

Selective utilization of medicinal polysaccharides by human gut Bacteroides and Parabacteroides species

Human gut Bacteroides and Parabacteroides species play crucial roles in human health and are known for their capacity to utilize diverse polysaccharides. Understanding how these bacteria utilize medicinal polysaccharides is foundational for developing polysaccharides-based prebiotics and drugs. Here, we systematically mapped the utilization profiles of 20 different medicinal polysaccharides by 28 human gut Bacteroides and Parabacteroides species. The growth profiles exhibited substantial variation across different bacterial species and medicinal polysaccharides. Ginseng polysaccharides promoted the growth of multiple Bacteroides and Parabacteroides species; in contrast, Dendrobium polysaccharides selectively promoted the growth of Bacteroides uniformis. This distinct utilization profile was associated with genomic variation in carbohydrate-active enzymes, rather than monosaccharides composition variation among medicinal polysaccharides. Through comparative transcriptomics and genetical manipulation, we validated that the polysaccharide utilization locus PUL34_Bu enabled Bacteroides uniformis to utilize Dendrobium polysaccharides (i.e. glucomannan). In addition, we found that the GH26 enzyme in PUL34_Bu allowed Bacteroides uniformis to utilize multiple plant-derived mannan. Overall, our results revealed the selective utilization of medicinal polysaccharide by Bacteroides and Parabacteroides species and provided insights into the use of polysaccharides in engineering the human gut microbiome.

Theoretical Study of Sphingomyelinases from Entamoeba histolytica and Trichomonas vaginalis Sheds Light on the Evolution of Enzymes Needed for Survival and Colonization

The path to survival for pathogenic organisms is not straightforward. Pathogens require a set of enzymes for tissue damage generation and to obtain nourishment, as well as a toolbox full of alternatives to bypass host defense mechanisms. Our group has shown that the parasitic protist Entamoeba histolytica encodes for 14 sphingomyelinases (SMases); one of them (acid sphingomyelinase 6, aSMase6) is involved in repairing membrane damage and exhibits hemolytic activity. The enzymatic characterization of aSMase6 has been shown to be activated by magnesium ions but not by zinc, as shown for the human aSMase, and is strongly inhibited by cobalt. However, no structural data are available for the aSMase6 enzyme. In this work, bioinformatic analyses showed that the protist aSMases are diverse enzymes, are evolutionarily related to hemolysins derived from bacteria, and showed a similar overall structure as parasitic, free-living protists and mammalian enzymes. AlphaFold3 models predicted the occupancy of cobalt ions in the active site of the aSMase6 enzyme. Cavity blind docking showed that the substrate is pushed outward of the active site when cobalt is bound instead of magnesium ions. Additionally, the structural models of the aSMase6 of E. histolytica showed a loop that is absent from the rest of the aSMases, suggesting that it may be involved in hemolytic activity, as demonstrated experimentally using the recombinant proteins of aSMase4 and aSMase6. Trichomonas vaginalis enzymes show a putative transmembrane domain and seem functionally different from E. histolytica. This work provides insight into the future biochemical analyses that can show mechanistic features of parasitic protists sphingomyelinases, ultimately rendering these enzymes potential therapeutic targets.

Applicability of AlphaFold2 in the modeling of dimeric, trimeric, and tetrameric coiled-coil domains

Coiled coils are a common protein structural motif involved in cellular functions ranging from mediating protein-protein interactions to facilitating processes such as signal transduction or regulation of gene expression. They are formed by two or more alpha helices that wind around a central axis to form a buried hydrophobic core. Various forms of coiled-coil bundles have been reported, each characterized by the number, orientation, and degree of winding of the constituent helices. This variability is underpinned by short sequence repeats that form coiled coils and whose properties determine both their overall topology and the local geometry of the hydrophobic core. The strikingly repetitive sequence has enabled the development of accurate sequence-based coiled-coil prediction methods; however, the modeling of coiled-coil domains remains a challenging task. In this work, we evaluated the accuracy of AlphaFold2 in modeling coiled-coil domains, both in modeling local geometry and in predicting global topological properties. Furthermore, we show that the prediction of the oligomeric state of coiled-coil bundles can be achieved by using the internal representations of AlphaFold2, with a performance better than any previous state-of-the-art method (code available at https://github.com/labstructbioinf/dc2_oligo).

Deciphering the cleavage sites of 3C-like protease in Gammacoronaviruses and Deltacoronaviruses

Coronaviruses replicate by using the 3C-like protease (3CLpro) to cleave polyprotein precursors and host proteins. However, current tools for identifying 3CLpro cleavage sites are limited, particularly in Gammacoronaviruses (GammaCoV) and Deltacoronaviruses (DeltaCoV). This study aims to fill this gap by identifying 3CLpro cleavage sites in these viruses to provide deeper insights into their pathogenic mechanisms. By integrating sequence alignments and structural model comparisons, we developed a position-specific scoring matrix (PSSM) based on self-cleavage motifs, revealing specific preferences for each residue. Utilizing AlphaFold2's predicted alignment error (PAE) and predicted local distance difference test (pLDDT), we found that most cleavage sequences are located in regions with high PAE and low pLDDT values. KEGG pathway analysis showed that potential host protein cleavage targets are mainly concentrated in pathways related to nucleo-cytoplasmic transport and endocytosis. Through in vitro cleavage experiments and mutational analysis, we identified and validated three high-scoring proteins-nucleoporin 58 (NUP58), cell division cycle 73 (CDC73), and signal transducing adaptor molecule 2 (STAM2). These findings suggest that 3CLpro not only plays a vital role in viral replication but may also influence host cell functions by cleaving host proteins. This study provides an effective tool for identifying 3CLpro cleavage sites, revealing the pathogenic mechanisms of coronaviruses, and offering new insights for developing potential therapeutic targets.

Target protein identification in live cells and organisms with a non-diffusive proximity tagging system

Identifying target proteins for bioactive molecules is essential for understanding their mechanisms, developing improved derivatives, and minimizing off-target effects. Despite advances in target identification (target-ID) technologies, significant challenges remain, impeding drug development. Most target-ID methods use cell lysates, but maintaining an intact cellular context is vital for capturing specific drug-protein interactions, such as those with transient protein complexes and membrane-associated proteins. To address these limitations, we developed POST-IT (Pup-On-target for Small molecule Target Identification Technology), a non-diffusive proximity tagging system for live cells, orthogonal to the eukaryotic system. POST-IT utilizes an engineered fusion of proteasomal accessory factor A and HaloTag to transfer Pup to proximal proteins upon directly binding to the small molecule. After significant optimization to eliminate self-pupylation and polypupylation, minimize depupylation, and optimize chemical linkers, POST-IT successfully identified known targets and discovered a new binder, SEPHS2, for dasatinib, and VPS37C as a new target for hydroxychloroquine, enhancing our understanding these drugs' mechanisms of action. Furthermore, we demonstrated the application of POST-IT in live zebrafish embryos, highlighting its potential for broad biological research and drug development.