lDDT: a local superposition-free score for comparing protein structures and models using distance difference tests

Analysis of enzyme kinetics of fungal methionine synthases in an optimized colorimetric microscale assay for measuring cobalamin-independent methionine synthase activity

Aspergillus spp. and Rhizopus spp., used in solid-state plant food fermentations, encode cobalamin-independent methionine synthase activity (MetE, EC 2.1.1.14). Here, we examine the enzyme kinetics, reaction activation energies (Ea), thermal robustness, and structural folds of three MetEs from three different food-fermentation relevant fungi, Aspergillus sojae, Rhizopus delemar, and Rhizopus microsporus, and compare them to the MetE from Escherichia coli. We also downscaled and optimized a colorimetric assay to allow direct MetE activity measurements in microplates. The catalytic rates, kcat, of the three fungal MetE enzymes on the methyl donor (6S)-5-methyl-tetrahydropteroyl-L-glutamate3 ranged from 1.2 to 3.3 min-1 and KM values varied from 0.8 to 6.8 µM. The kcat was lowest for the R. delemar MetE, but this enzyme also had the lowest KM thus resulting in the highest kcat/KM of ∼1.4 min-1 µM-1 among the three fungal enzymes. The kcat was higher for the E. coli enzyme, 12 min-1, but KM was 6.4 µM, resulting in kcat/KM of ∼1.9 min-1 µM-1. The Ea values of the fungal MetEs ranged from 52 to 97 kJ mole-1 and were higher than that of the E. coli MetE (38.7 kJ mole -1). The predicted structural folds of the MetEs were very similar. Tm values of the fungal MetEs ranged from 41 to 54 °C, highest for the A. sojae enzyme (54 °C), lowest for the R. delemar (41 °C). At 30 °C, the half-lives of the three fungal enzymes varied significantly, with MetE from A. sojae having the longest (> 600 min, kD=0), and R. delemar the shortest (17 min). Knowledge of the kinetics of these enzymes is important for understanding methionine synthesis in fungi and a first step in promoting methionine synthesis in fungally fermented plant foods.

Structure Bioinformatics of Six Human Integral Transmembrane Enzymes and their AlphaFold3 Predicted Water-Soluble QTY Analogs: Insights into FACE1 and STEA4 Binding Mechanisms

OBJECTIVE

Human integral membrane enzymes are essential for catalyzing a wide range of biochemical reactions and regulating key cellular processes. However, studying these enzymes remains challenging due to their hydrophobic nature, which necessitates the use of detergents. This study explores whether applying the QTY code can reduce the hydrophobicity of these enzymes while preserving their structures and functions, thus facilitating bioinformatics analysis of six key integral membrane enzymes: MGST2, LTC4S, PTGES, FACE1, STEA4, and SCD.

METHODS

The water-soluble QTY analogs of the six membrane enzymes were predicted using AlphaFold3. The predicted structures were superposed with CyroEM determined native structures in PyMOL to observe changes in structure and protein-ligand binding ability.

RESULTS

The native membrane enzymes superposed well with their respective QTY analogs, with the root mean square deviation (RMSD) ranging from 0.273 Å to 0.875 Å. Surface hydrophobic patches on the QTY analogs were significantly reduced. Importantly, the protein-ligand interactions in FACE1 and STEA4 were largely preserved, indicating maintained functionality.

CONCLUSION

Our structural bioinformatics studies using the QTY code and AlphaFold3 not only provide the opportunities of designing more water-soluble integral membrane enzymes, but also use these water-soluble QTY analogs as antigens for therapeutic monoclonal antibody discovery to specifically target the key integral membrane enzymes.

AlphaFold and what is next: bridging functional, systems and structural biology

INTRODUCTION

The DeepMind's AlphaFold (AF) has revolutionized biomedical and biocience research by providing both experts and non-experts with an invaluable tool for predicting protein structures. However, while AF is highly effective for predicting structures of rigid and globular proteins, it is not able to fully capture the dynamics, conformational variability, and interactions of proteins with ligands and other biomacromolecules.

AREAS COVERED

In this review, we present a comprehensive overview of the latest advancements in 3D model predictions for biomacromolecules using AF. We also provide a detailed analysis its of strengths and limitations, and explore more recent iterations, modifications, and practical applications of this strategy. Moreover, we map the path forward for expanding the landscape of AF toward predicting structures of every protein and peptide, and their interactions in the proteome in the most physiologically relevant form. This discussion is based on an extensive literature search performed using PubMed and Google Scholar.

EXPERT OPINION

While significant progress has been made to enhance AF's modeling capabilities, we argue that a combined approach integrating both various in silico and in vitro methods will be most beneficial for the future of structural biology, bridging the gaps between static and dynamic features of proteins and their functions.

Exploring miR-34a, miR-449, and ADAM2/ADAM7 Expressions as Potential Biomarkers in Male Infertility: A Combined In Silico and Experimental Approach

miR-34a and miR-449 are key miRNAs involved in sperm function and male fertility, with their dysregulation potentially contributing to male infertility. ADAM proteins, specifically ADAM2 and ADAM7, are also implicated in sperm function. This study investigates the interactions between miR-34a, miR-449, and ADAM2/ADAM7, exploring their roles in male infertility through both experimental analyses and molecular docking. In this case-control study, 15 infertile males and 15 healthy controls were included. Gene expression levels of miR-34a, miR-449, and SOX30 were measured using real-time PCR, while protein levels of ADAM7 and ADAM2 in sperm were assessed through western blotting. Additionally, molecular docking was performed to analyze the binding affinities between miR-34a/miR-449 and ADAM2/ADAM7, with docking scores and confidence levels evaluated. Expression levels of ADAM7 and ADAM2 proteins in sperm from the infertile group showed significant differences compared with the control group (P ≤ 0.05). A significant difference was observed in the expression of miR-449, miR-34a, and SOX30 genes between the control and infertile groups (P < 0.05). A significant correlation between miR-34a expression, ADAM7 protein expression, and sperm morphology was observed. However, no statistically significant correlation was found between miR-34a expression and sperm motility, sperm count, blastocyst, or embryo rates in ICSI and IVF (P ≥ 0.05). Molecular docking and dynamics studies revealed strong interactions between miR-34a/miR-449 and ADAM proteins. The ADAM7/miR-34a complex showed the highest binding affinity with a docking score of - 372.40 and a confidence score of 0.9884, followed by ADAM7/miR-449. Hydrogen bond analysis indicated stable binding, with 9 bonds for ADAM2/miR-34a and 7 for ADAM7/miR-34a. These interactions suggest a significant role in regulating sperm morphology and function.miR-34a, miR-449, ADAM7, and ADAM2 protein expression appear to be involved in the molecular mechanisms of male infertility. These parameters show potential as biomarkers in assisted reproductive technology techniques, particularly by influencing sperm morphology and function.

Sam-Sam Association Between EphA2 and SASH1: In Silico Studies of Cancer-Linked Mutations

Recently, SASH1 has emerged as a novel protein interactor of a few Eph tyrosine kinase receptors like EphA2. These interactions involve the first N-terminal Sam (sterile alpha motif) domain of SASH1 (SASH1-Sam1) and the Sam domain of Eph receptors. Currently, the functional meaning of the SASH1-Sam1/EphA2-Sam complex is unknown, but EphA2 is a well-established and crucial player in cancer onset and progression. Thus, herein, to investigate a possible correlation between the formation of the SASH1-Sam1/EphA2-Sam complex and EphA2 activity in cancer, cancer-linked mutations in SASH1-Sam1 were deeply analyzed. Our research plan relied first on searching the COSMIC database for cancer-related SASH1 variants carrying missense mutations in the Sam1 domain and then, through a variety of bioinformatic tools and molecular dynamic simulations, studying how these mutations could affect the stability of SASH1-Sam1 alone, leading eventually to a defective fold. Next, through docking studies, with the support of AlphaFold2 structure predictions, we investigated if/how mutations in SASH1-Sam1 could affect binding to EphA2-Sam. Our study, apart from presenting a solid multistep research protocol to analyze structural consequences related to cancer-associated protein variants with the support of cutting-edge artificial intelligence tools, suggests a few mutations that could more likely modulate the interaction between SASH1-Sam1 and EphA2-Sam.

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies

Genome foundation models hold transformative potential for precision medicine, drug discovery, and understanding complex biological systems. However, existing models are often inefficient, constrained by suboptimal tokenization and architectural design, and biased toward reference genomes, limiting their representation of low-abundance, uncultured microbes in the rare biosphere. To address these challenges, we developed GenomeOcean , a 4-billion-parameter generative genome foundation model trained on over 600 Gbp of high-quality contigs derived from 220 TB of metagenomic datasets collected from diverse habitats across Earth's ecosystems. A key innovation of GenomeOcean is training directly on large-scale co-assemblies of metagenomic samples, enabling enhanced representation of rare microbial species and improving generalizability beyond genome-centric approaches. We implemented a byte-pair encoding (BPE) tokenization strategy for genome sequence generation, alongside architectural optimizations, achieving up to 150× faster sequence generation while maintaining high biological fidelity. GenomeOcean excels in representing microbial species and generating protein-coding genes constrained by evolutionary principles. Additionally, its fine-tuned model demonstrates the ability to discover novel biosynthetic gene clusters (BGCs) in natural genomes and perform zero-shot synthesis of biochemically plausible, complete BGCs. GenomeOcean sets a new benchmark for metagenomic research, natural product discovery, and synthetic biology, offering a robust foundation for advancing these fields.

Functional profiling of the sequence stockpile: a protein pair-based assessment of in silico prediction tools

MOTIVATION

In silico functional annotation of proteins is crucial to narrowing the sequencing-accelerated gap in our understanding of protein activities. Numerous function annotation methods exist, and their ranks have been growing, particularly so with the recent deep learning-based developments. However, it is unclear if these tools are truly predictive. As we are not aware of any methods that can identify new terms in functional ontologies, we ask if they can, at least, identify molecular functions of proteins that are non-homologous to or far-removed from known protein families.

RESULTS

Here, we explore the potential and limitations of the existing methods in predicting the molecular functions of thousands of such proteins. Lacking the "ground truth" functional annotations, we transformed the assessment of function prediction into evaluation of functional similarity of protein pairs that likely share function but are unlike any of the currently functionally annotated sequences. Notably, our approach transcends the limitations of functional annotation vocabularies, providing a means to assess different-ontology annotation methods. We find that most existing methods are limited to identifying functional similarity of homologous sequences and fail to predict the function of proteins lacking reference. Curiously, despite their seemingly unlimited by-homology scope, deep learning methods also have trouble capturing the functional signal encoded in protein sequence. We believe that our work will inspire the development of a new generation of methods that push boundaries and promote exploration and discovery in the molecular function domain.

AVAILABILITY AND IMPLEMENTATION

The data underlying this article are available at https://doi.org/10.6084/m9.figshare.c.6737127.v3. The code used to compute siblings is available openly at https://bitbucket.org/bromberglab/siblings-detector/.

Molecular mechanism targeting condensin for chromosome condensation

Genomes are organised into DNA loops by the Structural Maintenance of Chromosomes (SMC) proteins. SMCs establish functional chromosomal sub-domains for DNA repair, gene expression and chromosome segregation, but how SMC activity is specifically targeted is unclear. Here, we define the molecular mechanism targeting the condensin SMC complex to specific chromosomal regions in budding yeast. A conserved pocket on the condensin HAWK subunit Ycg1 binds to chromosomal receptors carrying a related motif, CR1. In early mitosis, CR1 motifs in receptors Sgo1 and Lrs4 recruit condensin to pericentromeres and rDNA, to facilitate sister kinetochore biorientation and rDNA condensation, respectively. We additionally find that chromosome arm condensation begins as sister kinetochores come under tension, in a manner dependent on the Ycg1 pocket. We propose that multiple CR1-containing proteins recruit condensin to chromosomes and identify several additional candidates based on their sequence. Overall, we uncover the molecular mechanism that targets condensin to functionalise chromosomal domains to achieve accurate chromosome segregation during mitosis.

GDFold2: A fast and parallelizable protein folding environment with freely defined objective functions

An important step of mainstream protein structure prediction is to model the 3D protein structure based on the predicted 2D inter-residue geometric information. This folding step has been integrated into a unified neural network to allow end-to-end training in state-of-the-art methods like AlphaFold2, but is separately implemented using the Rosetta folding environment in some traditional methods like trRosetta. Despite the inferiority in prediction accuracy, the conventional approach allows for the sampling of various protein conformations compatible with the predicted geometric constraints, partially capturing the dynamic information. Here, we propose GDFold2, a novel protein folding environment, to address the limitations of Rosetta. On the one hand, GDFold2 is highly computationally efficient, capable of accomplishing multiple folding processes in parallel within the time scale of minutes for generic proteins. On the other hand, GDFold2 supports freely defined objective functions to fulfill diversified optimization requirements. Moreover, we propose a quality assessment (QA) model to provide reliable prediction on the quality of protein structures folded by GDFold2, thus substantially simplifying the selection of structural models. GDFold2 and the QA model could be combined to investigate the transition path between protein conformational states, and the online server is available at https://structpred.life.tsinghua.edu.cn/server_gdfold2.html.

Structural prediction of chimeric immunogen candidates to elicit targeted antibodies against betacoronaviruses

Betacoronaviruses pose an ongoing pandemic threat. Antigenic evolution of the SARS-CoV-2 virus has shown that much of the spontaneous antibody response is narrowly focused rather than broadly neutralizing against even SARS-CoV-2 variants, let alone future threats. One way to overcome this is by focusing the antibody response against better-conserved regions of the viral spike protein. This has been demonstrated empirically in prior work, but we posit that systematic design tools will further potentiate antigenic focusing approaches. Here, we present a design approach to predict stable chimeras between SARS-CoV-2 and other coronaviruses, creating synthetic spike proteins that display a desired conserved region, in this case S2, and vary other regions. We leverage AlphaFold to predict chimeric structures and create a new metric for scoring chimera stability based on AlphaFold outputs. We evaluated 114 candidate spike chimeras using this approach. Top chimeras were further evaluated using molecular dynamics simulation as an intermediate validation technique, showing good stability compared to low-scoring controls. Experimental testing of five predicted-stable and two predicted-unstable chimeras confirmed 5/7 predictions, with one intermediate result. This demonstrates the feasibility of the underlying approach, which can be used to design custom immunogens to focus the immune response against a desired viral glycoprotein epitope.

Visualizing gaussian-chain like structural models of human α-synuclein in monomeric pre-fibrillar state: Solution SAXS data and modeling analysis

Here, using small angle X-ray scattering (SAXS) data profile as reference, we attempted to visualize conformational ensemble accessible prefibrillar monomeric state of α-synuclein in solution. In agreement with previous reports, our analysis also confirmed that α-synuclein molecules adopted disordered shape profile under non-associating conditions. Chain-ensemble modeling protocol with dummy residues provided two weighted averaged clusters of semi-extended shapes. Further, Ensemble Optimization Method (EOM) computed mole fractions of semi-extended "twisted" conformations which might co-exist in solution. Since these were only Cα traces of the models, ALPHAFOLD2 server was used to search for all-atom models. Comparison with experimental data showed all predicted models disagreed equally, as individuals. Finally, we employed molecular dynamics simulations and normal mode analysis-based search coupled with SAXS data to seek better agreeing models. Overall, our analysis concludes that a shifting equilibrium of curved models with low α-helical content best-represents non-associating monomeric α-synuclein.

RNA-Puzzles Round V: blind predictions of 23 RNA structures

RNA-Puzzles is a collective endeavor dedicated to the advancement and improvement of RNA three-dimensional structure prediction. With agreement from structural biologists, RNA structures are predicted by modeling groups before publication of the experimental structures. We report a large-scale set of predictions by 18 groups for 23 RNA-Puzzles: 4 RNA elements, 2 Aptamers, 4 Viral elements, 5 Ribozymes and 8 Riboswitches. We describe automatic assessment protocols for comparisons between prediction and experiment. Our analyses reveal some critical steps to be overcome to achieve good accuracy in modeling RNA structures: identification of helix-forming pairs and of non-Watson-Crick modules, correct coaxial stacking between helices and avoidance of entanglements. Three of the top four modeling groups in this round also ranked among the top four in the CASP15 contest.

AFFIPred: AlphaFold2 structure-based Functional Impact Prediction of missense variations

Protein structure holds immense potential for pathogenicity prediction, albeit structure-based predictors are limited compared to the sequence-based counterparts due to the "structure knowledge gap" between large number of available protein sequences and relatively limited number of structures. Leveraging the highly accurate protein structures predicted by AlphaFold2 (AF2), we introduce AFFIPred, an ensemble machine learning classifier that combines sequence and AF2-based structural characteristics to predict missense variant pathogenicity. Based on the assessments on unseen datasets, AFFIPred reached a comparable level of performance with the state-of-the-art predictors such as AlphaMissense. We also showed that the recruitment of AF2 structures that are full-length and represent the unbound states ensures more precise SASA calculations compared to the recruitment of experimental structures. In line with the completeness of the AF2 structures, their use provide a more comprehensive view of the structural characteristics of the missense variation datasets by capturing all variants. AFFIPred maintains high-level accuracy without the limitations of PDB-based classifiers. AFFIPred has predicted over 210 million variations of the human proteome, which are accessible at https://affipred.timucinlab.com/.

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/.

Novel lineage of anelloviruses with large genomes identified in dolphins

Anellovirus infections are ubiquitous in mammals but lack any clear disease association, suggesting a commensal virus-host relationship. Although anelloviruses have been identified in numerous mammalian hosts, their presence in members of the family Delphinidae has yet to be reported. Here, using a metagenomic approach, we characterize complete anellovirus genomes (n = 69) from four Delphinidae host species: short-finned pilot whale (Globicephala macrorhynchus, n = 19), killer whale (Orcinus orca, n = 9), false killer whale (Pseudorca crassidens, n = 6), and pantropical spotted dolphin (Steno attenuatus, n = 1). Sequence comparison of the open reading frame 1 (ORF1) encoding the capsid protein, the only conserved gene shared by all anelloviruses, shows that the Delphinidae anelloviruses form a novel genus-level clade that encompasses 22 unique species-level groupings. We provide evidence that different Delphinidae species can be co-infected by multiple anelloviruses belonging to distinct species groupings. Notably, the ORF1 protein of the Delphinidae anelloviruses is considerably larger than those encoded by all previously described anelloviruses from other hosts (spanning 14 vertebrate orders and including 27 families). Comprehensive analysis of the ORF1 sequences and predicted protein structures showed that the increased size of these proteins results from divergent elaborations within the capsid-distal P2 subdomain and elongation of the C-terminal domain of ORF1. Comparative structural and phylogenetic analyses suggest that acquisition of the P2 subdomain and its diversification occurred convergently in the anelloviruses associated with primate and Delphinidae hosts. Collectively, our results further the appreciation of diversity and evolution of the ubiquitous and enigmatic viruses in the family Anelloviridae.

IMPORTANCE

Anelloviruses are ubiquitous in mammals, but their infection has not yet been linked to any disease, suggesting a commensal virus-host relationship. Here, we describe the first anelloviruses associated with diverse species of dolphins. The dolphinid anelloviruses represent a new genus (tentatively named "Qoptorquevirus") and encode open reading frame 1 (ORF1) (capsid) proteins that are considerably larger than those encoded by previously described anelloviruses from other hosts. Comprehensive analysis of the ORF1 sequences and predicted protein structures revealed the underlying structural basis for such an extravagant ORF1 size and suggested that ORF1 size increased convergently in the anelloviruses associated with primate and Delphinidae hosts, respectively. Collectively, our results provide insights into the diversity and evolution of Anelloviridae. Further exploration of the anellovirus diversity, especially in the host species that have not yet been sampled, is expected to further clarify their evolutionary trajectory and explain the unusual virus-host commensal relationship.

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.

AlphaFold 2, but not AlphaFold 3, predicts confident but unrealistic β-solenoid structures for repeat proteins

AlphaFold 2 (AF2) has revolutionised protein structure prediction but, like any new tool, its performance on specific classes of targets, especially those potentially under-represented in its training data, merits attention. Prompted by a highly confident prediction for a biologically meaningless, randomly permuted repeat sequence, we assessed AF2 performance on sequences composed of perfect repeats of random sequences of different lengths. AF2 frequently folds such sequences into β-solenoids which, while ascribed high confidence, contain unusual and implausible features such as internally stacked and uncompensated charged residues. A number of sequences confidently predicted as β-solenoids are predicted by other advanced methods as intrinsically disordered. The instability of some predictions is demonstrated by molecular dynamics. Importantly, other deep learning-based structure prediction tools predict different structures or β-solenoids with much lower confidence suggesting that AF2 alone has an unreasonable tendency to predict confident but unrealistic β-solenoids for perfect repeat sequences. The potential implications for structure prediction of natural (near-)perfect sequence repeat proteins are also explored.

Physical-aware model accuracy estimation for protein complex using deep learning method

With the breakthrough of AlphaFold2 on monomers, the research focus of structure prediction has shifted to protein complexes, driving the continued development of new methods for multimer structure prediction. Therefore, it is crucial to accurately estimate quality scores for the multimer model independent of the used prediction methods. In this work, we propose a physical-aware deep learning method, DeepUMQA-PA, to evaluate the residue-wise quality of protein complex models. Given the input protein complex model, the residue-based contact area and orientation features were first constructed using Voronoi tessellation, representing the potential physical interactions and hydrophobic properties. Then, the relationship between local residues and the overall complex topology as well as the inter-residue evolutionary information are characterized by geometry-based features, protein language model embedding representation, and knowledge-based statistical potential features. Finally, these features are fed into a fused network architecture employing equivalent graph neural network and ResNet network to estimate residue-wise model accuracy. Experimental results on the CASP15 test set demonstrate that our method outperforms the state-of-the-art method DeepUMQA3 by 3.69 % and 3.49 % on Pearson and Spearman, respectively. Notably, our method achieved 16.8 % and 15.5 % improvement in Pearson and Spearman, respectively, for the evaluation of nanobody-antigens. In addition, DeepUMQA-PA achieved better MAE scores than AlphaFold-Multimer and AlphaFold3 self-assessment methods on 43 % and 50 % of the targets, respectively. All these results suggest that physical-aware information based on the area and orientation of atom-atom and atom-solvent contacts has the potential to capture sequence-structure-quality relationships of proteins, especially in the case of flexible proteins. The DeepUMQA-PA server is freely available at http://zhanglab-bioinf.com/DeepUMQA-PA/.

NuFold: end-to-end approach for RNA tertiary structure prediction with flexible nucleobase center representation

RNA plays a crucial role not only in information transfer as messenger RNA during gene expression but also in various biological functions as non-coding RNAs. Understanding mechanical mechanisms of function needs tertiary structure information; however, experimental determination of three-dimensional RNA structures is costly and time-consuming, leading to a substantial gap between RNA sequence and structural data. To address this challenge, we developed NuFold, a novel computational approach that leverages state-of-the-art deep learning architecture to accurately predict RNA tertiary structures. NuFold is a deep neural network trained end-to-end for the output structure from the input sequence. NuFold incorporates a nucleobase center representation, which enables flexible conformation of ribose rings. Benchmark study showed that NuFold clearly outperformed energy-based methods and demonstrated comparable results with existing state-of-the-art deep-learning-based methods. NuFold exhibited a particular advantage in building correct local geometries of RNA. Analyses of individual components in the NuFold pipeline indicated that the performance improved by utilizing metagenome sequences for multiple sequence alignment and increasing the number of recycling. NuFold is also capable of predicting multimer complex structures of RNA by linking the input sequences.

Borrelia burgdorferi tolerates alteration to P66 porin function in a murine infectivity model

Borrelia burgdorferi exists in a complex enzootic life cycle requiring differential gene regulation. P66, a porin and adhesin, is upregulated and essential during mammalian infection, but is not produced or required within the tick vector. We sought to determine whether the porin function of P66 is essential for infection. Vancomycin treatment of B. burgdorferi cultures was used to screen for P66 porin function and found to generate spontaneous mutations in p66 (bb0603). Three novel, spontaneous, missense P66 mutants (G175V, T176M, and G584R) were re-created by site-directed mutagenesis in an infectious strain background and tested for infectivity in mice by ID50 experiments. Two of the three mutants retained infectivity comparable to the isogenic control, suggesting that B. burgdorferi can tolerate alteration to P66 porin function during infection. The third mutant exhibited highly attenuated infectivity and produced low levels of P66 protein. Interestingly, four isolates that were recovered for p66 sequencing from mouse tissues revealed novel secondary point mutations in genomic p66. However, these secondary mutations did not rescue P66 porin function. New structural modeling of P66 is presented and consistent with these experimental results. This is the first work to assess the contribution of P66 porin function to B. burgdorferi pathogenesis.

Integrated analysis of protein sequence and structure redefines viral diversity and the taxonomy of the Flaviviridae

The Flaviviridae are a family of non-segmented positive-sense enveloped RNA viruses containing significant pathogens including hepatitis C virus and yellow fever virus. Recent large-scale metagenomic surveys have identified many diverse RNA viruses related to classical orthoflaviviruses and pestiviruses but quite different genome lengths and configurations, and with a hugely expanded host range that spans multiple animal phyla, including molluscs, cnidarians and stramenopiles,, and plants. Grouping of RNA-directed RNA polymerase (RdRP) hallmark gene sequences of flavivirus and 'flavi-like' viruses into four divergent clades and multiple lineages within them was congruent with helicase gene phylogeny, PPHMM profile comparisons, and comparison of RdRP protein structure predicted by AlphFold2. These results support their classification into the established order, Amarillovirales, in three families (Flaviviridae, Pestiviridae, and Hepaciviridae), and 14 genera. This taxonomic framework informed by RdRP hallmark gene evolutionary relationships provides a stable reference from which major genome re-organisational events can be understood.

Gradations in protein dynamics captured by experimental NMR are not well represented by AlphaFold2 models and other computational metrics

The advent of accurate methods to predict the fold of proteins initiated by AlphaFold2 is rapidly changing our understanding of proteins and helping their design. However, these methods are mainly trained on protein structures determined with X-ray diffraction, where the protein is packed in crystals at often cryogenic temperatures. They can therefore only reliably cover well-folded parts of proteins that experience few, if any, conformational changes. Experimentally, solution nuclear magnetic resonance (NMR) is the experimental method of choice to gain insight into protein dynamics at near physiological conditions. Computationally, methods such as molecular dynamics (MD) simulations and Normal Mode Analysis (NMA) allow the estimation of a protein's intrinsic flexibility based on a single protein structure. This work addresses, on a large scale, the relationships for proteins between the AlphaFold2 pLDDT metric, the observed dynamics in solution from NMR metrics, interpreted MD simulations, and the computed dynamics with NMA from single AlphaFold2 models and NMR ensembles. We observe that these metrics agree well for rigid residues that adopt a single well-defined conformation, which are clearly distinct from residues that exhibit dynamic behavior and adopt multiple conformations. This direct order/disorder categorisation is reflected in the correlations observed between the parameters, but becomes very limited when considering only the likely dynamic residues. The gradations of dynamics observed by NMR in flexible protein regions are therefore not represented by these computational approaches. Our results are interactively available for each protein from https://bio2byte.be/af_nmr_nma/.

multistrap: boosting phylogenetic analyses with structural information

In a phylogeny, trustworthy reliability branch support estimates are as important as the tree itself. We show that reliability support values based on bootstrapping can be improved by combining sequence and structural information from proteins. Our approach relies on the systematic comparison of homologous intra-molecular structural distances. These variations exhibit less saturation than sequence-based Hamming distances and support the computation of tree-like distance matrices resolvable into phylogenetic trees using distance-based methods such as minimum evolution. These trees bear strong similarities to their sequence-based counterparts and allow the estimation of bootstrap support values, but they are sufficiently distinct so that their information content may be combined. The combined sequence and structure bootstrap support values yield improved discrimination between correct and incorrect branches. In this work we show that our approach, named multistrap, is suitable for the improvement of bootstrap branch support values using both predicted and experimental 3D structures.

AlphaFold2 and ESMFold: A large-scale pairwise model comparison of human enzymes upon Pfam functional annotation

AlphaFold2 predicts protein structures from structural and functional knowledge. Alternatively, ESMFold does the same adopting protein language models. Here, we map available Pfam domains on pairs of models of the human reference proteome computed with both procedures and we compare the mapped regions relevant for functional annotation. We find that, rather irrespectively of the global superimposition of the pairwise models, Pfam-containing regions overlap with a TM-score above 0.8 and a predicted local distance difference test (pLDDT) which is higher than the rest of the modeled sequence. This indicates that both methods are similarly performing in modeled regions that overlap Pfam domains, carrying structural and functional information, with pLDDT values slightly higher for AlphaFold2. The mapping of 9834 Pfam domains also allows the location of 2578 active sites in 3382 enzymes of the human proteome, including 807 proteins for which the active site is not reported in UniProt.

Deep learning-based assessment of missense variants in the COG4 gene presented with bilateral congenital cataract

OBJECTIVE

We compared the protein structure and pathogenicity of clinically relevant variants of the COG4 gene with AlphaFold2 (AF2), Alpha Missense (AM), and ThermoMPNN for the first time.

METHODS AND ANALYSIS

The sequences of clinically relevant Cog4 missense variants (one novel identified p.Y714F and three pre-existing p.G512R, p.R729W and p.L769R from Uniprot Q9H9E3) were imported into AF2 for protein structural prediction, and the pathogenicity was estimated using AM and ThermoMPNN. Different pathogenicity metrics were aggregated with principal component analysis (PCA) and further analysed at three levels (amino acid position, substitution and post-translation) based on all possible Cog4 missense variants (n=14 915).

RESULTS

Localised protein structural impact including change of conformation and amino acid polarity, breakage of hydrogen bond and salt-bridge, and formation of alpha-helix were identified among clinically relevant Cog4 variants. The global structural comparison with multidimensional scaling demonstrated variants with similar protein structures (AF2) tended to exhibit similar clinical and biological phenotypes. The Cog4 p.Y714F variant exhibited greater protein structural similarity to mutated Cog4 found in Saul‒Wilson syndrome (p.G512R) and shared similar clinical phenotype (congenital cataract and psychomotor retardation). PCA of included pathogenic metrics demonstrated p.Y714F occurred at a critical position in Cog4 amino acid sequence with disrupted post-translational phosphorylation.

CONCLUSION

Deep learning algorithms, including AF2, AM and ThermoMPNN, can be useful for evaluating variant of uncertain significance (VUS) by structural and pathogenicity prediction. Despite classified as VUS (American College of Medical Genetics and Genomics criteria: PM1, PP4), the pathogenicity in this Cog4 variant cannot be ruled out and warrants further investigation.

Engineered T cells secreting αB7-H3-αCD3 bispecific engagers for enhanced anti-tumor activity against B7-H3 positive multiple myeloma: a novel therapeutic approach

BACKGROUND

Multiple myeloma (MM) is an incurable plasma cell malignancy with increasing global incidence. Chimeric antigen receptor (CAR) T-cell therapy targeting BCMA has shown efficacy in relapsed or refractory MM, but it faces resistance due to antigen loss and the tumor microenvironment. Bispecific T-cell engaging (BITE) antibodies also encounter clinical challenges, including short half-lives requiring continuous infusion and potential toxicities.

METHODS

To address these issues, we developed a lentiviral system to engineer T cells that secrete αB7-H3-αCD3 bispecific engager molecules (αB7-H3-αCD3 ENG-T cells). We evaluated their effectiveness against MM cells with varying B7-H3 expression levels, from B7-H3neg to B7-H3high.

RESULTS

The αB7-H3-αCD3 ENG-T cells demonstrated significant anti-tumor activity against MM cell lines expressing B7-H3. SupT-1 cells (B7-H3neg) served as controls and exhibited minimal cytotoxicity from αB7-H3-αCD3 ENG T cells. In contrast, these engineered T cells showed dose-dependent killing of B7-H3-expressing MM cells: NCI-H929 (B7-H3low), L-363 (B7-H3medium), and KMS-12-PE (B7-H3high). For NCI-H929 cells, cytotoxicity reached 38.5 ± 7.4% (p = 0.0212) and 54.0 ± 9.2% (p = 0.0317) at effector-to-target (E:T) ratios of 5:1 and 10:1, respectively. Against L-363 cells, cytotoxicity was 56.6 ± 3.2% (p < 0.0001) and 71.4 ± 5.2% (p = 0.0002) at E:T ratios of 5:1 and 10:1, respectively. For KMS-12-PE cells, significant cytotoxic effects were observed even at an E:T ratio of 1:1, with 27.2 ± 3.7% (p = 0.0004), 44.4 ± 3.7% (p < 0.0001), and 68.6 ± 9.2% (p = 0.0004) cytotoxicity at E:T ratios of 1:1, 5:1, and 10:1, respectively.

CONCLUSIONS

These results indicate that αB7-H3-αCD3 ENG T cells could be a promising therapy for B7-H3-positive MM. They may enhance current MM treatments and improve overall outcomes. Additional preclinical and clinical research is required to fully assess their therapeutic potential.

Identification and functional analyses of drought stress resistance genes by transcriptomics of the Mongolian grassland plant Chloris virgata

BACKGROUND

Mongolian grasslands, including the Gobi Desert, have been exposed to drought conditions with few rains. In such harsh environments, plants with highly resistant abilities against drought stress survive over long periods. We hypothesized that these plants could harbor novel and valuable genes for enhancing drought stress resistance.

RESULTS

In this study, we identified Chloris virgata, a Mongolian grassland plant with strong drought resistance. RNA-seq-based transcriptome analysis was performed to uncover genes associated with drought stress resistance in C. virgata. De novo transcriptome assembly revealed 25,469 protein-coding transcripts and 1,219 upregulated genes after 3- and 6-hr drought stress treatments. Analysis by homology search and Gene Ontology (GO) enrichment indicated that abscisic acid (ABA)- and drought stress-related GO terms were enriched. Among the highly induced genes, ten candidate cDNAs were selected and overexpressed in Arabidopsis. When subjected to drought stress, three of these genes conferred strong drought resistance in the transgenic plants. We named these genes Mongolian Grassland plant Drought-stress resistance genes 1, 2, and 3 (MGD1, MGD2, and MGD3). Gene expression analyses in the transformants suggested that MGD1, MGD2, and MGD3 may activate drought stress-related signalling pathways.

CONCLUSION

This study highlighted the drought resistance of C. virgata and identified three novel genes that enhance drought stress resistance.

Synthetic T-Cell Receptor-like Protein Behaves as a Janus Particle in Solution

Protein engineering enables the creation of tailor-made proteins for a variety of applications. ImmTACs stand out as promising therapeutics for cancer and other treatments while also presenting unique challenges for stability, formulation, and delivery. We have shown that ImmTACs behave as Janus particles in solution, leading to self-association at low concentrations, even when the average protein-protein interactions suggest that the molecule should be stable. The formation of small but stable oligomers was confirmed by static and dynamic light scattering and analytical ultracentrifugation. Modeling of the structure using AlphaFold leads to a rational explanation for this behavior, consistent with the Janus particle assembly observed for inverse patchy particles.

Plasticity of BioPhi-driven humanness optimization in ScFv-CD99 binding affinity validated through AlphaFold, HADDOCK, and MD simulations

BioPhi-guided humanization was utilized to enhance the humanness of a humanized single-chain variable fragment targeting CD99, leading to the development of two variants: HuScFvMT99/3BP and HuScFvMT99/3HY. The HuScFvMT99/3BP variant incorporated framework region modifications, leading to modest improvements in humanness, particularly in the VH domain, although the VL domain remained suboptimal. To address this limitation, HuScFvMT99/3HY was designed by combining the VL domain of wild-type with the VH domain of HuScFvMT99/3BP. Molecular dynamics simulations employing AlphaFold2, AlphaFold3, and HADDOCK were performed to evaluate the HuScFv-CD99 peptide complexes. AF2-based simulations demonstrated enhanced binding free energy (ΔGbinding) for both variants compared to HuScFvMT99/3WT. However, ΔGbinding values obtained from AF3 and HD simulations were inconsistent, with HuScFvMT99/3BP exhibiting the weakest binding affinity. While ΔGbinding patterns derived from AlphaFold3 and HADDOCK simulations aligned, amino acid decomposition analysis revealed variations in the interaction coordinates of the predicted complexes. Root-mean-square deviation analysis indicated improved structural stability for HuScFvMT99/3BP (0.975 Å) and HuScFvMT99/3HY (1.075 Å) relative to HuScFvMT99/3WT (1.225 Å). Biolayer interferometry further confirmed that HuScFvMT99/3WT exhibited the highest binding affinity (KD = 1.35 × 10⁻⁷ M) compared to HuScFvMT99/3BP (KD = 2.64 × 10⁻⁷ M) and HuScFvMT99/3HY (KD = 3.95 × 10⁻⁷ M). Supporting evidence was provided by ELISA and flow cytometry experiments. PITHA analysis revealed a high immunogenicity risk for all variants, despite HuScFvMT99/3HY displaying improved humanness, a larger complementarity-determining region (CDR) cavity, and a more hydrophobic CDR-H3 loop. These findings highlight the delicate balance between enhancing humanness and preserving the structural and functional integrity critical for therapeutic antibody development.

The prevalence of motility-related genes within the human oral microbiota

The human oral and nasal microbiota contains approximately 770 cultivable bacterial species. More than 2,000 genome sequences of these bacteria can be found in the expanded Human Oral Microbiome Database (eHOMD). We developed HOMDscrape, a freely available Python software tool to programmatically retrieve and process amino acid sequences and sequence identifiers from BLAST results acquired from the eHOMD website. Using the data obtained through HOMDscrape, the phylogeny of proteins involved in bacterial type 9 secretion system (T9SS)-driven gliding motility, flagellar motility, and type IV pilus-driven twitching motility was constructed. A comprehensive phylogenetic analysis was conducted for all components of the rotary T9SS, a machinery responsible for secreting various enzymes, virulence factors, and enabling bacterial gliding motility. Results revealed that the T9SS outer membrane β-barrel protein SprA of human oral bacteria underwent horizontal evolution. Overall, we catalog motile bacteria that inhabit the human oral microbiota and document their evolutionary connections. These results will serve as a guide for further studies exploring the impact of motility on the shaping of the human oral microbiota.IMPORTANCEThe human oral microbiota has been extensively studied, and many of the isolated bacteria have genome sequences stored on the human oral microbiome database (eHOMD). Spatial distribution and polymicrobial biofilms are observed in the oral microbiota, but little is understood on how they are influenced by motility. To bridge this gap, we developed a software tool to identify motile bacteria from eHOMD. The results enabled the cataloging of motile bacteria present in the oral microbiota but also provided insight into their evolutionary relationships. This information can guide future research to better understand how bacterial motility shapes the human oral microbiota.

dbAMP 3.0: updated resource of antimicrobial activity and structural annotation of peptides in the post-pandemic era

Antimicrobial resistance is one of the most urgent global health threats, especially in the post-pandemic era. Antimicrobial peptides (AMPs) offer a promising alternative to traditional antibiotics, driving growing interest in recent years. dbAMP is a comprehensive database offering extensive annotations on AMPs, including sequence information, functional activity data, physicochemical properties and structural annotations. In this update, dbAMP has curated data from over 5200 publications, encompassing 33,065 AMPs and 2453 antimicrobial proteins from 3534 organisms. Additionally, dbAMP utilizes ESMFold to determine the three-dimensional structures of AMPs, providing over 30,000 structural annotations that facilitate structure-based functional insights for clinical drug development. Furthermore, dbAMP employs molecular docking techniques, providing over 100 docked complexes that contribute useful insights into the potential mechanisms of AMPs. The toxicity and stability of AMPs are critical factors in assessing their potential as clinical drugs. The updated dbAMP introduced an efficient tool for evaluating the hemolytic toxicity and half-life of AMPs, alongside an AMP optimization platform for designing AMPs with high antimicrobial activity, reduced toxicity and increased stability. The updated dbAMP is freely accessible at https://awi.cuhk.edu.cn/dbAMP/. Overall, dbAMP represents a comprehensive and essential resource for AMP analysis and design, poised to advance antimicrobial strategies in the post-pandemic era.

Fluorescence lifetime sorting reveals tunable enzyme interactions within cytoplasmic condensates

Ribonucleoprotein (RNP) condensates partition RNA and protein into multiple liquid phases. The multiphasic feature of condensate-enriched components creates experimental challenges for distinguishing membraneless condensate functions from the surrounding dilute phase. We combined fluorescence lifetime imaging microscopy (FLIM) with phasor plot filtering and segmentation to resolve condensates from the dilute phase. Condensate-specific lifetimes were used to track protein-protein interactions by measuring FLIM-Förster resonance energy transfer (FRET). We used condensate FLIM-FRET to evaluate whether mRNA decapping complex subunits can form decapping-competent interactions within P-bodies. Condensate FLIM-FRET revealed the presence of core subunit interactions within P-bodies under basal conditions and the disruption of interactions between the decapping enzyme (Dcp2) and a critical cofactor (Dcp1A) during oxidative stress. Our results show a context-dependent plasticity of the P-body interaction network, which can be rewired within minutes in response to stimuli. Together, our FLIM-based approaches provide investigators with an automated and rigorous method to uncover and track essential protein-protein interaction dynamics within RNP condensates in live cells.

Reliable protein-protein docking with AlphaFold, Rosetta, and replica-exchange

Despite the recent breakthrough of AlphaFold (AF) in the field of protein sequence-to-structure prediction, modeling protein interfaces and predicting protein complex structures remains challenging, especially when there is a significant conformational change in one or both binding partners. Prior studies have demonstrated that AF-multimer (AFm) can predict accurate protein complexes in only up to 43% of cases.1 In this work, we combine AlphaFold as a structural template generator with a physics-based replica exchange docking algorithm to better sample conformational changes. Using a curated collection of 254 available protein targets with both unbound and bound structures, we first demonstrate that AlphaFold confidence measures (pLDDT) can be repurposed for estimating protein flexibility and docking accuracy for multimers. We incorporate these metrics within our ReplicaDock 2.0 protocol2to complete a robust in-silico pipeline for accurate protein complex structure prediction. AlphaRED (AlphaFold-initiated Replica Exchange Docking) successfully docks failed AF predictions including 97 failure cases in Docking Benchmark Set 5.5. AlphaRED generates CAPRI acceptable-quality or better predictions for 63% of benchmark targets. Further, on a subset of antigen-antibody targets, which is challenging for AFm (20% success rate), AlphaRED demonstrates a success rate of 43%. This new strategy demonstrates the success possible by integrating deep-learning based architectures trained on evolutionary information with physics-based enhanced sampling. The pipeline is available at github.com/Graylab/AlphaRED.

An in silico framework to visualize how cancer-associated mutations influence structural plasticity of the chemokine receptor CCR3

G protein Coupled Receptors (GPCRs) are the largest family of cell surface receptors in humans. Somatic mutations in GPCRs are implicated in cancer progression and metastasis, but mechanisms are poorly understood. Emerging evidence implicates perturbation of intra-receptor activation pathway motifs whereby extracellular signals are transmitted intracellularly. Recently, sufficiently sensitive methodology was described to calculate structural strain as a function of missense mutations in AlphaFold-predicted model structures, which was extensively validated on experimental and predicted structural datasets. When paired with Molecular Dynamics (MD) simulations, these tools provide a facile approach to screen mutations in silico. We applied this framework to calculate the structural and dynamic effects of cancer-associated mutations in the chemokine receptor CCR3, a Class A GPCR involved in cancer and autoimmune disorders. Residue-residue contact scoring refined effective strain results, highlighting significant remodeling of inter- and intra-motif contacts along the highly conserved GPCR activation pathway network. We then integrated AlphaFold-derived predicted Local Distance Difference Test scores with per-residue Root Mean Square Fluctuations and activation pathway Contact Analysis (CONAN) from coarse grain MD simulations to identify statistically significant changes in receptor dynamics upon mutation. Finally, analysis of negative control mutants suggests false positive results in AlphaFold pipelines should be considered but can be mitigated with stricter control of statistical analysis. Our results indicate selected mutants influence structural plasticity of CCR3 related to ligand interaction, activation, and G protein coupling, using a framework that could be applicable to a wide range of biochemically relevant protein targets following further validation.

Insights into electron transfer and bifurcation of the Synechocystis sp. PCC6803 hydrogenase reductase module

The NAD+-reducing soluble [NiFe] hydrogenase (SH) is the key enzyme for production and consumption of molecular hydrogen (H2) in Synechocystis sp. PCC6803. In this study, we focused on the reductase module of the SynSH and investigated the structural and functional aspects of its subunits, particularly the so far elusive role of HoxE. We demonstrated the importance of HoxE for enzyme functionality, suggesting a regulatory role in maintaining enzyme activity and electron supply. Spectroscopic analysis confirmed that HoxE and HoxF each contain one [2Fe2S] cluster with an almost identical electronic structure. Structure predictions, alongside experimental evidence for ferredoxin interactions, revealed a remarkable similarity between SynSH and bifurcating hydrogenases, suggesting a related functional mechanism. Our study unveiled the subunit arrangement and cofactor composition essential for biological electron transfer. These findings enhance our understanding of NAD+-reducing [NiFe] hydrogenases in terms of their physiological function and structural requirements for biotechnologically relevant modifications.

The β2 adrenergic receptor cross-linked interactome identifies 14-3-3 proteins as regulating the availability of signaling-competent receptors

The emerging picture of G protein-coupled receptor function suggests that the global signaling response is an integrated sum of a multitude of individual receptor responses, each regulated by their local protein environment. The β2 adrenergic receptor (B2AR) has long served as an example receptor in the development of this model. However, the mechanism and the identity of the protein-protein interactions that govern the availability of receptors competent for signaling remain incompletely characterized. To address this question, we characterized the interactome of agonist-stimulated B2AR in human embryonic kidney 293 cells using FLAG coimmunoprecipitation coupled to stable isotope labeling by amino acids in cell culture and mass spectrometry. Our B2AR cross-linked interactome identified 190 high-confidence proteins, including almost all known interacting proteins and 6 out of 7 isoforms of the 14-3-3 family of scaffolding proteins. Inhibiting 14-3-3 proteins with the peptide difopein enhanced isoproterenol-stimulated adrenergic signaling via cAMP approximately 3-fold and increased both miniGs and arrestin recruitment to B2AR more than 2-fold each, without noticeably changing EC50 with respect to cAMP signaling or effector recruitment upon stimulation. Our results show that 14-3-3 proteins negatively regulate downstream signaling by inhibiting access of B2AR to effector proteins. We propose that 14-3-3 proteins maintain a dynamic pool of B2AR that has reduced signaling efficacy in response to acute agonist stimulation, limiting the number of signaling-competent receptors at the plasma membrane. SIGNIFICANCE STATEMENT: This study presents a new interactome of the agonist-stimulated β2 adrenergic receptor, a paradigmatic G protein-coupled receptor that is both a model system for members of this class and an important signaling protein in respiratory, cardiovascular, and metabolic regulation. We identify 14-3-3 proteins as responsible for restricting β2 adrenergic receptor access to signaling effectors and maintaining a receptor population that is insensitive to acute stimulation by agonists.

How the technologies behind self-driving cars, social networks, ChatGPT, and DALL-E2 are changing structural biology

The performance of deep Neural Networks (NNs) in the text (ChatGPT) and image (DALL-E2) domains has attracted worldwide attention. Convolutional NNs (CNNs), Large Language Models (LLMs), Denoising Diffusion Probabilistic Models (DDPMs)/Noise Conditional Score Networks (NCSNs), and Graph NNs (GNNs) have impacted computer vision, language editing and translation, automated conversation, image generation, and social network management. Proteins can be viewed as texts written with the alphabet of amino acids, as images, or as graphs of interacting residues. Each of these perspectives suggests the use of tools from a different area of deep learning for protein structural biology. Here, I review how CNNs, LLMs, DDPMs/NCSNs, and GNNs have led to major advances in protein structure prediction, inverse folding, protein design, and small molecule design. This review is primarily intended as a deep learning primer for practicing experimental structural biologists. However, extensive references to the deep learning literature should also make it relevant to readers who have a background in machine learning, physics or statistics, and an interest in protein structural biology.

Challenges in bridging the gap between protein structure prediction and functional interpretation

The rapid evolution of protein structure prediction tools has significantly broadened access to protein structural data. Although predicted structure models have the potential to accelerate and impact fundamental and translational research significantly, it is essential to note that they are not validated and cannot be considered the ground truth. Thus, challenges persist, particularly in capturing protein dynamics, predicting multi-chain structures, interpreting protein function, and assessing model quality. Interdisciplinary collaborations are crucial to overcoming these obstacles. Databases like the AlphaFold Protein Structure Database, the ESM Metagenomic Atlas, and initiatives like the 3D-Beacons Network provide FAIR access to these data, enabling their interpretation and application across a broader scientific community. Whilst substantial advancements have been made in protein structure prediction, further progress is required to address the remaining challenges. Developing training materials, nurturing collaborations, and ensuring open data sharing will be paramount in this pursuit. The continued evolution of these tools and methodologies will deepen our understanding of protein function and accelerate disease pathogenesis and drug development discoveries.

Comparison of structural networks across homologous proteins

Protein sequence determines its structure and function. The indirect relationship between protein function and structure lies deep-rooted in the structural topology that has evolved into performing optimal function. The evolution of structure and its interconnectivity has been conventionally studied by comparing the root means square deviation between protein structures at the backbone level. Two factors that are necessary for the quantitative comparison of non-covalent interactions are (a) explicit inclusion of the coordinates of side-chain atoms and (b) consideration of multiple structures from the conformational landscape to account for structural variability. We have recently addressed these fundamental issues by investigating the alteration of inter-residue interactions across an ensemble of protein structure networks through a graph spectral approach. In this study, we have developed a rigorous method to compare the structure networks of homologous proteins, with a wide range of sequence identity percentages. A range of dissimilarity measures that show the extent of change in the network across homologous structures are generated, which also includes the comparison of the protein structure variability. We discuss in detail, scenarios where the variation of structure is not accompanied by loss or gain of the overall network and its vice versa. The sequence-based phylogeny among the homologs is also compared with the lineage obtained from information from such a robust structure comparison. In summary, we can obtain a quantitative comparison score for the structure networks of homologous proteins, which also enables us to study the evolution of protein function based on the variation of their topologies.

Synthetic anti-RNA antibody derivatives for RNA visualization in mammalian cells

Although antibody derivatives, such as Fabs and scFvs, have revolutionized the cellular imaging, quantification and tracking of proteins, analogous tools and strategies are unavailable for cellular RNA visualization. Here, we developed four synthetic anti-RNA scFv (sarabody) probes and their green fluorescent protein (GFP) fusions and demonstrated their potential to visualize RNA in live mammalian cells. We expressed these sarabodies and sarabody-GFP modules, purified them as soluble proteins, characterized their binding interactions with their corresponding epitopes and finally employed two of the four modules, sara1-GFP and sara1c-GFP, to visualize a target messenger RNA in live U2OS cells. Our current RNA imaging strategy is analogous to the existing MCP-MS2 system for RNA visualization, but additionally, our approach provides robust flexibility for developing target RNA-specific imaging modules, as epitope-specific probes can be selected from a library generated by diversifying the sarabody complementarity determining regions. While we continue to optimize these probes, develop new probes for various target RNAs and incorporate other fluorescence proteins like mCherry and HaloTag, our groundwork results demonstrated that these first-of-a-kind immunofluorescent probes will have tremendous potential for tracking mature RNAs and may aid in visualizing and quantifying many cellular processes as well as examining the spatiotemporal dynamics of various RNAs.

RNA-TorsionBERT: leveraging language models for RNA 3D torsion angles prediction

MOTIVATION

Predicting the 3D structure of RNA is an ongoing challenge that has yet to be completely addressed despite continuous advancements. RNA 3D structures rely on distances between residues and base interactions but also backbone torsional angles. Knowing the torsional angles for each residue could help reconstruct its global folding, which is what we tackle in this work. This paper presents a novel approach for directly predicting RNA torsional angles from raw sequence data. Our method draws inspiration from the successful application of language models in various domains and adapts them to RNA.

RESULTS

We have developed a language-based model, RNA-TorsionBERT, incorporating better sequential interactions for predicting RNA torsional and pseudo-torsional angles from the sequence only. Through extensive benchmarking, we demonstrate that our method improves the prediction of torsional angles compared to state-of-the-art methods. In addition, by using our predictive model, we have inferred a torsion angle-dependent scoring function, called TB-MCQ, that replaces the true reference angles by our model prediction. We show that it accurately evaluates the quality of near-native predicted structures, in terms of RNA backbone torsion angle values. Our work demonstrates promising results, suggesting the potential utility of language models in advancing RNA 3D structure prediction.

AVAILABILITY AND IMPLEMENTATION

Source code is freely available on the EvryRNA platform: https://evryrna.ibisc.univ-evry.fr/evryrna/RNA-TorsionBERT.

PAbFold: Linear Antibody Epitope Prediction using AlphaFold2

Defining the binding epitopes of antibodies is essential for understanding how they bind to their antigens and perform their molecular functions. However, while determining linear epitopes of monoclonal antibodies can be accomplished utilizing well-established empirical procedures, these approaches are generally labor- and time-intensive and costly. To take advantage of the recent advances in protein structure prediction algorithms available to the scientific community, we developed a calculation pipeline based on the localColabFold implementation of AlphaFold2 that can predict linear antibody epitopes by predicting the structure of the complex between antibody heavy and light chains and target peptide sequences derived from antigens. We found that this AlphaFold2 pipeline, which we call PAbFold, was able to accurately flag known epitope sequences for several well-known antibody targets (HA / Myc) when the target sequence was broken into small overlapping linear peptides and antibody complementarity determining regions (CDRs) were grafted onto several different antibody framework regions in the single-chain antibody fragment (scFv) format. To determine if this pipeline was able to identify the epitope of a novel antibody with no structural information publicly available, we determined the epitope of a novel anti-SARS-CoV-2 nucleocapsid targeted antibody using our method and then experimentally validated our computational results using peptide competition ELISA assays. These results indicate that the AlphaFold2-based PAbFold pipeline we developed is capable of accurately identifying linear antibody epitopes in a short time using just antibody and target protein sequences. This emergent capability of the method is sensitive to methodological details such as peptide length, AlphaFold2 neural network versions, and multiple-sequence alignment database. PAbFold is available at https://github.com/jbderoo/PAbFold.

Evolution-based protein engineering: functional switching between transthyretins and 5-hydroxyisourate hydrolases

Transthyretin (TTR) is a vertebrate-exclusive transport protein that plays a key role in binding and distributing thyroid hormones. However, its evolutionary origin lies in the duplication of the gene that encoding the enzyme 5-hydroxyisourate hydrolase (HIUase), which is involved in uric acid metabolism. Unlike TTR, HIUase is ubiquitous in both prokaryotes and eukaryotes, with the exception of hominids. Both HIUase and TTR subfamilies form homotetramers, possessing an internal charged cavity between the two dimer pairs. Based on their high degree of structural similarity, we hypothesized that specific in silico substitutions would enable the interconversion between these protein functions. Using an evolution-based approach, we engineered two putative protein sequences, where correlated locally conserved positions from one subfamily representative sequence were substituted by the other, and vice versa. Applying computational modeling techniques, the best models were refined, validated, and their cavity volumes, three-dimensional geometries, propensity to aggregation and electrostatic potentials were analyzed. Molecular dynamics simulations were performed with the reference proteins and the engineered mutants in the bound and unbound states. We demonstrate that the volumes and geometries differ from one another, due to size and physicochemical differences between their ligands. The bound state mutant complexes are stable, and the enzymatic assay demonstrated active new enzymes. Our work suggests that the evolution-based protein engineering approach used has residue-specific resolution to identify locally conserved residues in the sequence of evolutionarily related proteins, such as HIUase and TTR.

Exploring the Structure-Function Relationships in a 5-Aminolevulinic Acid Synthase and the Use of Protein Engineering to Expand its Substrate Range

5-Aminolevulinate synthase (ALAS) is a PLP-dependent enzyme that catalyzes the production of 5-aminolevulinate from succinyl-CoA and glycine. Its ability to catalyze the essentially irreversible C-C bond formation has significant potential in chemoenzymatic synthesis of α-amino ketones. Native ALAS, unfortunately, is extremely substrate-selective, and this seriously limits its synthetic utility. Here, we have used three different protein engineering strategies to overcome this problem for the acyl-CoA substrate. By combining previously reported mutation results and structural analysis, a series of site-saturation mutagenesis/screening efforts were focused on R21, T82, N84, and T362 of Rhodopseudomonas palustris ALAS. These yielded single, double, and triple mutants with significantly improved substrate ranges. The steady-state kinetic parameters of several key variants were determined. These data were analyzed in the framework of the ALAS catalytic mechanism to identify the steps that may have been impacted. The most active variant was used in a larger-scale reaction to demonstrate its synthetic potential. Taken together, our results show how ALAS might become a useful biocatalyst for α-amino ketone synthesis and have also allowed us to comment on the relative merits of each the three protein engineering strategies utilized.

Dissecting Cytophagalysin: Structural and Biochemical Studies of a Bacterial Pappalysin-Family Metallopeptidase

Cytophaga is a genus of Gram-negative bacteria occurring in soil and the gut microbiome. It is closely related to pathogenic Flavobacterium spp. that cause severe diseases in fish. Cytophaga strain L43-1 secretes cytophagalysin (CPL1), a 137 kDa peptidase with reported collagenolytic and gelatinolytic activity. We performed highly-confident structure prediction calculations for CPL1, which identified 11 segments and domains, including a signal peptide for secretion, a prosegment (PS) for latency, a metallopeptidase (MP)-like catalytic domain (CD), and eight immunoglobulin (Ig)-like domains (D3-D10). In addition, two short linkers were found at the D8-D9 and D9-D10 junctions, and the structure would be crosslinked by four disulfide bonds. The CPL1 CD was found closest to ulilysin from Methanosarcina acetivorans, which assigns CPL1 to the lower-pappalysin family within the metzincin clan of MPs. Based on the structure predictions, we aimed to produce constructs spanning the full-length enzyme, as well as PS+CD, PS+CD+D3, and PS+CD+D3+D4. However, we were successful only with the latter three constructs. We could activate recombinant CPL1 by PS removal employing trypsin, and found that both zymogen and mature CPL1 were active in gelatin zymography and against a fluorogenic gelatin variant. This activity was ablated in a mutant, in which the catalytic glutamate described for lower pappalyins and other metzincins was replaced by alanine, and by a broad-spectrum metal chelator. Overall, these results proved that our recombinant CPL1 is a functional active MP, thus supporting the conclusions derived from the structure predictions.

Phosphatidylinositol 4,5-bisphosphate drives the formation of EGFR and EphA2 complexes

Receptor tyrosine kinases (RTKs) regulate many cellular functions and are important targets in pharmaceutical development, particularly in cancer treatment. EGFR and EphA2 are two key RTKs that are associated with oncogenic phenotypes. Several studies have reported functional interplay between these receptors, but the mechanism of interaction is still unresolved. Here, we use a time-resolved fluorescence spectroscopy called PIE-FCCS to resolve EGFR and EphA2 interactions in live cells. We tested the role of ligands and found that EGF, but not ephrin A1 (EA1), stimulated heteromultimerization between the receptors. To determine the effect of anionic lipids, we targeted phospholipase C (PLC) activity to alter the abundance of phosphatidylinositol 4,5-bisphosphate (PIP2). We found that higher PIP2 levels increased homomultimerization of both EGFR and EphA2, as well as heteromultimerization. This study provides a direct characterization of EGFR and EphA2 interactions in live cells and shows that PIP2 can have a substantial effect on the spatial organization of RTKs.

In Silico design of a multi-epitope vaccine for Human Parechovirus: Integrating immunoinformatics and computational techniques

Human parechovirus (HPeV) is widely recognized as a severe viral infection affecting infants and neonates. Belonging to the Picornaviridae family, HPeV is categorized into 19 distinct genotypes. Among them, HPeV-1 is the most prevalent genotype, primarily associated with respiratory and digestive symptoms. Considering HPeV's role as a leading cause of life-threatening viral infections in infants and the lack of effective antiviral therapies, our focus centered on developing two multi-epitope vaccines, namely HPeV-Vax-1 and HPeV-Vax-2, using advanced immunoinformatic techniques. Multi-epitope vaccines have the advantage of protecting against various virus strains and may be preferable to live attenuated vaccines. Using the NCBI database, three viral protein sequences (VP0, VP1, and VP3) from six HPeV strains were collected to construct consensus protein sequences. Then the antigenicity, toxicity, allergenicity, and stability were analyzed after discovering T-cell and linear B-cell epitopes from the protein sequences. The fundamental structures of the vaccines were produced by fusing the selected epitopes with appropriate linkers and adjuvants. Comprehensive physicochemical, antigenic, allergic assays, and disulfide engineering demonstrated the effectiveness of the vaccines. Further refinement of secondary and tertiary models for both vaccines revealed promising interactions with toll-like receptor 4 (TLR4) in molecular docking, further confirmed by molecular dynamics simulation. In silico immunological modeling was employed to assess the vaccine's capacity to stimulate an immune reaction. In silico immunological simulations were employed to evaluate the vaccines' ability to trigger an immune response. Codon optimization and in silico cloning analyses showed that Escherichia coli (E. coli) was most likely the host for the candidate vaccines. Our findings suggest that these multi-epitope vaccines could be the potential HPeV vaccines and are recommended for further wet-lab investigation.

Rational design of monomeric IL37 variants guided by stability and dynamical analyses of IL37 dimers

IL37 plays important roles in the regulation of innate immunity and its oligomeric status is critical to these roles. In its monomeric state, IL37 can effectively inhibit the inflammatory response of IL18 by binding to IL18Rα, a capacity lost in its dimeric form, underlining the pivotal role of the oligomeric status of IL37 in its anti-inflammatory action. Until now, two IL37 dimer structures have been deposited in PDB, reflecting a substantial difference in their dimer interfaces. Given this discrepancy, we analyzed the PDB structures of the IL37 dimer (PDB IDs: 6ncu, 5hn1) along with a AF2-multimer prediction by molecular dynamics (MD) simulations. Results showed that the 5hn1 and AF2-predicted dimers have the same interface and stably maintained their conformations throughout simulations, while the recent IL37 dimer (PDB ID: 6ncu) with a different interface did not, proposing a possible issue with the recent IL37 dimer structure (6ncu). Next, focusing on the stable dimer structures, we have identified five critical positions of V71/Y85/I86/E89/S114, three new positions compared to the literature, that would reduce dimer stability without affecting the monomer structure. Two quintuple mutants were tested by MD simulations and showed partial or complete dissociation of the dimer. Overall, the insights gained from this study reinforce the validity of the 5hn1 and AF2 multimer structures, while also advancing our understanding of the IL37 dimer interface through the generation of monomer-locked IL37 variants.

Accurate RNA 3D structure prediction using a language model-based deep learning approach

Accurate prediction of RNA three-dimensional (3D) structures remains an unsolved challenge. Determining RNA 3D structures is crucial for understanding their functions and informing RNA-targeting drug development and synthetic biology design. The structural flexibility of RNA, which leads to the scarcity of experimentally determined data, complicates computational prediction efforts. Here we present RhoFold+, an RNA language model-based deep learning method that accurately predicts 3D structures of single-chain RNAs from sequences. By integrating an RNA language model pretrained on ~23.7 million RNA sequences and leveraging techniques to address data scarcity, RhoFold+ offers a fully automated end-to-end pipeline for RNA 3D structure prediction. Retrospective evaluations on RNA-Puzzles and CASP15 natural RNA targets demonstrate the superiority of RhoFold+ over existing methods, including human expert groups. Its efficacy and generalizability are further validated through cross-family and cross-type assessments, as well as time-censored benchmarks. Additionally, RhoFold+ predicts RNA secondary structures and interhelical angles, providing empirically verifiable features that broaden its applicability to RNA structure and function studies.

Recent advances and challenges in protein complex model accuracy estimation

Estimation of model accuracy plays a crucial role in protein structure prediction, aiming to evaluate the quality of predicted protein structure models accurately and objectively. This process is not only key to screening candidate models that are close to the real structure, but also provides guidance for further optimization of protein structures. With the significant advancements made by AlphaFold2 in monomer structure, the problem of single-domain protein structure prediction has been widely solved. Correspondingly, the importance of assessing the quality of single-domain protein models decreased, and the research focus has shifted to estimation of model accuracy of protein complexes. In this review, our goal is to provide a comprehensive overview of the reference and statistical metrics, as well as representative methods, and the current challenges within four distinct facets (Topology Global Score, Interface Total Score, Interface Residue-Wise Score, and Tertiary Residue-Wise Score) in the field of complex EMA.