Protein Structure Prediction Research Articles

African Swine Fever Virus Protein-Protein Interaction Prediction.

The African swine fever virus (ASFV) is an often deadly disease in swine and poses a threat to swine livestock and swine producers. With its complex genome containing more than 150 coding regions, developing effective vaccines for this virus remains a challenge due to a lack of basic knowledge about viral protein function and protein-protein interactions between viral proteins and between viral and host proteins. In this work, we identified ASFV-ASFV protein-protein interactions (PPIs) using artificial intelligence-powered protein structure prediction tools. We benchmarked our PPI identification workflow on the Vaccinia virus, a widely studied nucleocytoplasmic large DNA virus, and found that it could identify gold-standard PPIs that have been validated in vitro in a genome-wide computational screening. We applied this workflow to more than 18,000 pairwise combinations of ASFV proteins and were able to identify seventeen novel PPIs, many of which have corroborating experimental or bioinformatic evidence for their protein-protein interactions, further validating their relevance. Two protein-protein interactions, I267L and I8L, I267L__I8L, and B175L and DP79L, B175L__DP79L, are novel PPIs involving viral proteins known to modulate host immune response.

Conformational Dynamics and Molecular Characterization of Alsin MORN Monomer and Dimeric Assemblies.

Despite the ubiquity of membrane occupation recognition nexus (MORN) motifs across diverse species in both eukaryotic and prokaryotic organisms, these protein domains remain poorly characterized. Their significance is underscored in the context of the Alsin protein, implicated in the debilitating condition known as infantile-onset ascending hereditary spastic paralysis (IAHSP). Recent investigations have proposed that mutations within the Alsin MORN domain disrupt proper protein assembly, precluding the formation of the requisite tetrameric configuration essential for the protein's inherent biological activity. However, a comprehensive understanding of the relationship between the biological functions of Alsin and its three-dimensional molecular structure is hindered by the lack of available experimental structures. In this study, we employed and compared several protein structure prediction algorithms to identify a three-dimensional structure for the putative MORN of Alsin. Furthermore, inspired by experimental pieces of evidence from previous studies, we employed the developed models to predict and investigate two homo-dimeric assemblies, characterizing their stability. This study's insights into the three-dimensional structure of the Alsin MORN domain and the stability dynamics of its homo-dimeric assemblies suggest an antiparallel linear configuration stabilized by a noncovalent interaction network.

Prediction of protein secondary structure by the improved TCN-BiLSTM-MHA model with knowledge distillation

Secondary structure prediction is a key step in understanding protein function and biological properties and is highly important in the fields of new drug development, disease treatment, bioengineering, etc. Accurately predicting the secondary structure of proteins helps to reveal how proteins are folded and how they function in cells. The application of deep learning models in protein structure prediction is particularly important because of their ability to process complex sequence information and extract meaningful patterns and features, thus significantly improving the accuracy and efficiency of prediction. In this study, a combined model integrating an improved temporal convolutional network (TCN), bidirectional long short-term memory (BiLSTM), and a multi-head attention (MHA) mechanism is proposed to enhance the accuracy of protein prediction in both eight-state and three-state structures. One-hot encoding features and word vector representations of physicochemical properties are incorporated. A significant emphasis is placed on knowledge distillation techniques utilizing the ProtT5 pretrained model, leading to performance improvements. The improved TCN, achieved through multiscale fusion and bidirectional operations, allows for better extraction of amino acid sequence features than traditional TCN models. The model demonstrated excellent prediction performance on multiple datasets. For the TS115, CB513 and PDB (2018–2020) datasets, the prediction accuracy of the eight-state structure of the six datasets in this paper reached 88.2%, 84.9%, and 95.3%, respectively, and the prediction accuracy of the three-state structure reached 91.3%, 90.3%, and 96.8%, respectively. This study not only improves the accuracy of protein secondary structure prediction but also provides an important tool for understanding protein structure and function, which is particularly applicable to resource-constrained contexts and provides a valuable tool for understanding protein structure and function.

Comparative analysis of thermal adaptations of extremophilic prolyl oligopeptidases

Prolyl oligopeptidases from psychrophilic, mesophilic, and thermophilic organisms found in a range of natural environments are studied using a combination of protein structure prediction, atomistic molecular dynamics, and trajectory analysis to determine how the S9 protease family adapts to extreme thermal conditions. We compare our results with hypotheses from the literature regarding structural adaptations that allow proteins to maintain structure and function at extreme temperatures, and we find that, in the case of prolyl oligopeptidases, only a subset of proposed adaptations are employed for maintaining stability. The catalytic and propeller domains are highly structured, limiting the range of mutations that can be made to enhance hydrophobicity or form disulfide bonds without disrupting the formation of necessary secondary structure. Rather, we observe a pattern in which overall prevalence of bound interactions (salt bridges and hydrogen bonds) is conserved by using increasing numbers of increasingly short-lived interactions as temperature increases. This suggests a role for an entropic rather than energetic strategy for thermal adaptation in this protein family.

Evaluating the 3D structure prediction tools to identify optimal MEBPVC structure models

Multi-Epitope based Peptide Vaccines Candidates (MEBPVCs) are peptide sequences with the immunogenic epitopes interspersed with linkers, adjuvants and other components. The optimization of one of the components of MEBPVCs, epitope order, resulted in a unique epitope ordered MEBPVC showing improved vaccine potency. In addition to the above combinatorial optimization task, predicting the 3D structure of these MEBPVCs presents a significant challenge owing to the diverse modeling approaches employed by the modeling tools. Here we embarked on evaluating the performance of the threading based ab initio modeling tool: I-TASSER, Large Language Model (LLM) based tool: ESMFold and the MSA based artificial intelligence based tool: AlphaFold2. Our investigation employed RMSD, PEA, SASA, Solubility, Globularity, secondary structure analysis, Docking, MDS and MVP. I-TASSER emerged as the superior performer. The analysis revealed a strong correlation between predicted tertiary structure and secondary structure composition, especially the β-turns and random coils. It appears that accurate secondary structure identification to be a pivotal step, dictating the propagation of accuracy (or inaccuracy). This finding highlights the critical importance of secondary structure prediction/assignment algorithms for enhanced protein structure modeling accuracy especially to integrate predictions where evolutionary information is minimal and or lacking. Small sample size is attributed to the low significance of the parameters though Globularity and SASA have shown good p-values. Overall, the results support the need and utilization of the LLMs in biological research, underscoring their potential to revolutionize protein structure prediction and advance diverse biomolecular applications, accelerating Synthetic Biology research.

EFG-CS: Predicting chemical shifts from amino acid sequences with protein structure prediction using machine learning and deep learning models.

Nuclear magnetic resonance (NMR) crystallography is one of the main methods in structural biology for analyzing protein stereochemistry and structure. The chemical shift of the resonance frequency reflects the effect of the protons in a molecule producing distinct NMR signals in different chemical environments. Apprehending chemical shifts from NMR signals can be challenging since having an NMR structure does not necessarily provide all the required chemical shift information, making predictive models essential for accurately deducing chemical shifts, either from protein structures or, more ideally, directly from amino acid sequences. Here, we present EFG-CS, a web server that specializes in chemical shift prediction. EFG-CS employs a machine learning-based transfer prediction model for backbone atom chemical shift prediction, using ESMFold-predicted protein structures. Additionally, ESG-CS incorporates a graph neural network-based model to provide comprehensive side-chain atom chemical shift predictions. Our method demonstrated reliable performance in backbone atom prediction, achieving comparable accuracy levels with root mean square errors (RMSE) of 0.30 ppm for H, 0.22 ppm for Hα, 0.89 ppm for C, 0.89 ppm for Cα, 0.84 ppm for Cβ, and 1.69 ppm for N. Moreover, our approach also showed predictive capabilities in side-chain atom chemical shift prediction achieving RMSE values of 0.71 ppm for Hβ, 0.74-1.15 ppm for Hδ, and 0.58-0.94 ppm for Hγ, solely utilizing amino acid sequences without homology or feature curation. This work shows for the first time that generative AI protein models can predict NMR shifts nearly comparable to experimental models. This web server is freely available at https://biosig.lab.uq.edu.au/efg_cs, and the chemical shift prediction results can be downloaded in tabular format and visualized in 3D format.

Comparative Analysis Highlights Uniconazole's Efficacy in Enhancing the Cold Stress Tolerance of Mung Beans by Targeting Photosynthetic Pathways.

Soybean (Glycine max) and mung bean (Vigna radiata) are key legumes with global importance, but their mechanisms for coping with cold stress-a major challenge in agriculture-have not been thoroughly investigated, especially in a comparative study. This research aimed to fill this gap by examining how these two major legumes respond differently to cold stress and exploring the role of uniconazole, a potential stress mitigator. Our comprehensive approach involved transcriptomic and metabolomic analyses, revealing distinct responses between soybean and mung bean under cold stress conditions. Notably, uniconazole was found to significantly enhance cold tolerance in mung bean by upregulating genes associated with photosynthesis, while its impact on soybean was either negligible or adverse. To further understand the molecular interactions, we utilized advanced machine learning algorithms for protein structure prediction, focusing on photosynthetic pathways. This enabled us to identify LOC106780309 as a direct binding target for uniconazole, confirmed through isothermal titration calorimetry. This research establishes a new comparative approach to explore how soybean and mung bean adapt to cold stress, offers key insights to improve the hardiness of legumes against environmental challenges, and contributes to sustainable agricultural practices and food security.

SpSIZ1 from hyperaccumulator Sedum plumbizincicola orchestrates SpABI5 to fine-tune cadmium tolerance.

Sedum plumbizincicola is a renowned hyperaccumulator of cadmium (Cd), possesses significant potential for eco-friendly phytoremediation of soil contaminated with Cd. Nevertheless, comprehension of the mechanisms underpinning its Cd stress response remains constrained, primarily due to the absence of a comprehensive genome sequence and an established genetic transformation system. In this study, we successfully identified a novel protein that specifically responds to Cd stress through early comparative iTRAQ proteome and transcriptome analyses under Cd stress conditions. To further investigate its structure, we employed AlphaFold, a powerful tool for protein structure prediction, and found that this newly identified protein shares a similar structure with Arabidopsis AtSIZ1. Therefore, we named it Sedum plumbizincicola SIZ1 (SpSIZ1). Our study revealed that SpSIZ1 plays a crucial role in positively regulating Cd tolerance through its coordination with SpABI5. Overexpression of SpSIZ1 significantly enhanced plant resistance to Cd stress and reduced Cd accumulation. Expression pattern analysis revealed higher levels of SpSIZ1 expression in roots compared to stems and leaves, with up-regulation under Cd stress induction. Importantly, overexpressing SpSIZ1 resulted in lower Cd translocation factors (Tfs) but maintained relatively constant Cd levels in roots under Cd stress, leading to enhanced Cd stress resistance in plants. Protein interaction analysis revealed that SpSIZ1 interacts with SpABI5, and the expression of genes responsive to abscisic acid (ABA) through SpABI5-dependent signaling was significantly up-regulated in SpSIZ1-overexpressing plants with Cd stress treatment. Collectively, our results illustrate that SpSIZ1 interacts with SpABI5, enhancing the expression of ABA downstream stress-related genes through SpABI5, thereby increasing Cd tolerance in plants.

Synthesis and Semi-Synthesis of Alpha-Synuclein: Insight into the Chemical Complexity of Synucleinopathies.

The chemical rules governing protein folding have intrigued generations of researchers for decades. With the advent of artificial intelligence (AI), prediction of protein structure has improved tremendously. However, there is still a level of analysis that is only possible through wet laboratory experiments, especially in respect to the investigation of the pathological effect of mutations and posttranslational modifications (PTMs) on proteins of interest. This requires the availability of pure peptides and proteins in sufficient quantities for biophysical, biochemical, and functional studies. In this context, chemical protein synthesis and semi-synthesis are powerful tools in protein research, which help to enlighten the role of protein modification in the physiology and pathology of proteins. A protein of high interest in the field of biomedicine is alpha-synuclein (aSyn), a protein deeply associated with several devastating neurodegenerative disorders such as Parkinson's disease (PD), dementia with Lewy bodies (DLB), or multiple systems atrophy (MSA). Here, we describe several methods and pathways to synthesize native or modified aSyn, and discuss how these approaches enable us to address pathological mechanisms that may open novel perspectives for therapeutic intervention.

Structural and Functional Characterization of New Lipid Transfer Proteins with Chitin-Binding Properties: Insights from Protein Structure Prediction, Molecular Docking, and Antifungal Activity.

Faced with the emergence of multiresistant microorganisms that affect human health, microbial agents have become a serious global threat, affecting human health and plant crops. Antimicrobial peptides have attracted significant attention in research for the development of new microbial control agents. This work's goal was the structural characterization and analysis of antifungal activity of chitin-binding peptides from Capsicum baccatum and Capsicum frutescens seeds on the growth of Candida and Fusarium species. Proteins were initially submitted to extraction in phosphate buffer pH 5.4 and subjected to chitin column chromatography. Posteriorly, two fractions were obtained for each species, Cb-F1 and Cf-F1 and Cb-F2 and Cf-F2, respectively. The Cb-F1 (C. baccatum) and Cf-F1 (C. frutescens) fractions did not bind to the chitin column. The electrophoresis results obtained after chromatography showed two major protein bands between 3.4 and 14.2 kDa for Cb-F2. For Cf-F2, three major bands were identified between 6.5 and 14.2 kDa. One band from each species was subjected to mass spectrometry, and both bands showed similarity to nonspecific lipid transfer protein. Candida albicans and Candida tropicalis had their growth inhibited by Cb-F2. Cf-F2 inhibited the development of C. albicans but did not inhibit the growth of C. tropicalis. Both fractions were unable to inhibit the growth of Fusarium species. The toxicity of the fractions was tested in vivo on Galleria mellonella larvae, and both showed a low toxicity rate at high concentrations. As a result, the fractions have enormous promise for the creation of novel antifungal compounds.

AggreProt: a web server for predicting and engineering aggregation prone regions in proteins.

Recombinant proteins play pivotal roles in numerous applications including industrial biocatalysts or therapeutics. Despite the recent progress in computational protein structure prediction, protein solubility and reduced aggregation propensity remain challenging attributes to design. Identification of aggregation-prone regions is essential for understanding misfolding diseases or designing efficient protein-based technologies, and as such has a great socio-economic impact. Here, we introduce AggreProt, a user-friendly webserver that automatically exploits an ensemble of deep neural networks to predict aggregation-prone regions (APRs) in protein sequences. Trained on experimentally evaluated hexapeptides, AggreProt compares to or outperforms state-of-the-art algorithms on two independent benchmark datasets. The server provides per-residue aggregation profiles along with information on solvent accessibility and transmembrane propensity within an intuitive interface with interactive sequence and structure viewers for comprehensive analysis. We demonstrate AggreProt efficacy in predicting differential aggregation behaviours in proteins on several use cases, which emphasize its potential for guiding protein engineering strategies towards decreased aggregation propensity and improved solubility. The webserver is freely available and accessible at https://loschmidt.chemi.muni.cz/aggreprot/.

PANDA-3D: protein function prediction based on AlphaFold models.

Previous protein function predictors primarily make predictions from amino acid sequences instead of tertiary structures because of the limited number of experimentally determined structures and the unsatisfying qualities of predicted structures. AlphaFold recently achieved promising performances when predicting protein tertiary structures, and the AlphaFold protein structure database (AlphaFold DB) is fast-expanding. Therefore, we aimed to develop a deep-learning tool that is specifically trained with AlphaFold models and predict GO terms from AlphaFold models. We developed an advanced learning architecture by combining geometric vector perceptron graph neural networks and variant transformer decoder layers for multi-label classification. PANDA-3D predicts gene ontology (GO) terms from the predicted structures of AlphaFold and the embeddings of amino acid sequences based on a large language model. Our method significantly outperformed a state-of-the-art deep-learningmethod that was trained with experimentally determined tertiary structures, and either outperformed or was comparable with several other language-model-based state-of-the-art methods with amino acid sequences as input.PANDA-3D is tailored to AlphaFold models, and the AlphaFold DB currently contains over 200 million predicted protein structures (as of May 1st, 2023), making PANDA-3D a useful tool that can accurately annotate the functions of a large number of proteins. PANDA-3D can be freely accessed as a web server from http://dna.cs.miami.edu/PANDA-3D/ and as a repository from https://github.com/zwang-bioinformatics/PANDA-3D.

Review and Comparative Analysis of Methods and Advancements in Predicting Protein Complex Structure.

Protein complexes perform diverse biological functions, and obtaining their three-dimensional structure is critical to understanding and grasping their functions. In many cases, it's not just two proteins interacting to form a dimer; instead, multiple proteins interact to form a multimer. Experimentally resolving protein complex structures can be quite challenging. Recently, there have been efforts and methods that build upon prior predictions of dimer structures to attempt to predict multimer structures. However, in comparison to monomeric protein structure prediction, the accuracy of protein complex structure prediction remains relatively low. This paper provides an overview of recent advancements in efficient computational models for predicting protein complex structures. We introduce protein-protein docking methods in detail and summarize their main ideas, applicable modes, and related information. To enhance prediction accuracy, other critical protein-related information is also integrated, such as predicting interchain residue contact, utilizing experimental data like cryo-EM experiments, and considering protein interactions and non-interactions. In addition, we comprehensively reviewcomputational approaches for end-to-end prediction ofprotein complex structuresbased on artificial intelligence (AI) technologyand describe commonly used datasets and representative evaluation metrics in protein complexes. Finally, we analyze the formidable challenges faced in current protein complex structure prediction tasks, including the structure prediction of heteromeric complex, disordered regions in complex, antibody-antigen complex, and RNA-related complex, as well as the evaluation metrics for complex assessment. We hope that this work will provide comprehensive knowledge of complex structure predictions to contribute to future advanced predictions.

ITPR1 variant-induced autosomal dominant hereditary spastic paraplegia in a Chinese family.

Hereditary spastic paraplegia (HSP) is a rare neurodegenerative disease prominently characterized by slowly progressive lower limb weakness and spasticity. The significant genotypic and phenotypic heterogeneity of this disease makes its accurate diagnosis challenging. In this study, we identified the NM_001168272: c.2714A > G (chr3.hg19: g.4716912A > G, N905S) variant in the ITPR1 gene in a three-generation Chinese family with multiple individuals affected by HSP, which we believed to be associated with HSP pathogenesis. To confirm, we performed whole exome sequencing, copy number variant assays, dynamic mutation analysis of the entire family, and protein structure prediction. The variant identified in this study was in the coupling domain, and this is the first corroborated report assigning ITPR1 variants to HSP. These findings expand the clinical and genetic spectrum of HSP and provide important data for its genetic analysis and diagnosis.

Efficient protein structure archiving using ProteStAr.

The introduction of Deep Minds' Alpha Fold 2 enabled the prediction of protein structures at an unprecedented scale. AlphaFold Protein Structure Database and ESM Metagenomic Atlas contain hundreds of millions of structures stored in CIF and/or PDB formats. When compressed with a general-purpose utility like gzip, this translates to tens of terabytes of data, which hinders the effective use of predicted structures in large-scale analyses. Here, we present ProteStAr, a compressor dedicated to CIF/PDB, as well as supplementary PAE files. Its main contribution is a novel approach to predicting atom coordinates on the basis of the previously analyzed atoms. This allows efficient encoding of the coordinates, the largest component of the protein structure files. The compression is lossless by default, though the lossy mode with a controlled maximum error of coordinates reconstruction is also present. Compared to the competing packages, i.e. BinaryCIF, Foldcomp, PDC, our approach offers a superior compression ratio at established reconstruction accuracy. By the efficient use of threads at both compression and decompression stages, the algorithm takes advantage of the multicore architecture of current central processing units and operates with speeds of about 1 GB/s. The presence of Python and C++ API further increases the usability of the presented method. The source code of ProteStAr is available at https://github.com/refresh-bio/protestar.

LazyAF, a pipeline for accessible medium-scale in silico prediction of protein-protein interactions.

Artificial intelligence has revolutionized the field of protein structure prediction. However, with more powerful and complex software being developed, it is accessibility and ease of use rather than capability that is quickly becoming a limiting factor to end users. LazyAF is a Google Colaboratory-based pipeline which integrates the existing ColabFold BATCH software to streamline the process of medium-scale protein-protein interaction prediction. LazyAF was used to predict the interactome of the 76 proteins encoded on the broad-host-range multi-drug resistance plasmid RK2, demonstrating the ease and accessibility the pipeline provides.

Using Attention-UNet Models to Predict Protein Contact Maps.

Proteins are essential to life, and understanding their intrinsic roles requires determining their structure. The field of proteomics has opened up new opportunities by applying deep learning algorithms to large databases of solved protein structures. With the availability of large data sets and advanced machine learning methods, the prediction of protein residue interactions has greatly improved. Protein contact maps provide empirical evidence of the interacting residue pairs within a protein sequence. Template-free protein structure prediction systems rely heavily on this information. This article proposes UNet-CON, an attention-integrated UNet architecture, trained to predict residue-residue contacts in protein sequences. With the predicted contacts being more accurate than state-of-the-art methods on the PDB25 test set, the model paves the way for the development of more powerful deep learning algorithms for predicting protein residue interactions.

3D STRUCTURE AND FUNCTION ANALYSIS OF RECOMBINANT ALDII PROTEIN FROM Uncultured Acidilobus sp. USING I-TASSER

The AldII protein is a new recombinant protein produced from a novel gene obtained via a metagenome approach. Previous studies showed that this protein is strong and has the same metal binding aspect as the class II Aldolase enzyme. Aldolase is a valuable enzyme used in pharmaceuticals, food processing, and biochemistry. Further investigation is required to comprehend the structure and function of the AldII protein due to its potential. Researchers will conduct sophisticated bioinformatic analysis on the 3D shape and function prediction of AldII using the I-TASSER webserver from Zhanglab. The AldII protein is a new recombinant protein produced from a novel gene obtained via a metagenome approach. Previous studies showed that this protein is strong and has the same metal binding aspect as the class II Aldolase enzyme. Aldolase is a valuable enzyme used in pharmaceuticals, food processing, and biochemistry. Further investigation is required to comprehend the structure and function of the AldII protein due to its potential. Researchers will conduct sophisticated bioinformatic analysis on the 3D shape and function prediction of AldII using the I-TASSER webserver from Zhanglab. The I-TASSER server is an online tool for the automated prediction of protein structure and annotation of functions based on structure. Analysis of the AldII protein using the I-TASSER webserver shows that this protein has a stable structure with the closest structural homology to deoxyribose-phosphate aldolase from Bacillus thuringiensis with PDB code 6btdA. Additionally, the biological structure analysis shows that this protein shares the biological function of the enzyme L-fuculose-1-phosphate aldolase, which is part of the class II Aldolase enzyme that plays a role in the catabolism of arabinose, L-Fuculose, and Rhamnose. The results align with prior research that states the AldII protein is a stable protein with a catalytic side that is homologous to the class II Aldolase enzyme.

The TIR1/AFB Family in Solanum melongena: Genome-Wide Identification and Expression Profiling under Stresses and Picloram Treatment

TIR1/AFB proteins are a class of auxin receptors with key roles in plant development and biotic and abiotic stress responses; several have been identified as targets of the auxin-mimicking herbicide picloram. In this study, we identified five putative TIR1/AFB gene family members in the important vegetable crop Solanum melongena (eggplant) and characterized them using bioinformatics tools and gene expression analyses. Phylogenetic analysis of the TIR1/AFBs classified them into three subgroups based on their Arabidopsis and Solanum lycopersicum homologs. AFB6 homologs were present only in S. melongena and S. lycopersicum, whereas AFB2/3 homologs were found only in Arabidopsis. One pair of S. melongena TIR1 homologs were located in syntenic regions in the genome and appeared to have arisen by segmental duplication. Promoter analysis revealed 898 cis-elements in the TIR1/AFB promoters, 125 of which were related to hormones, stress, light, or growth responses, but only SmAFB5 had a cis-acting regulatory element involved in auxin responsiveness (AuxRR-core). RNA sequencing and expression profiling showed that the TIR1/AFB genes were differentially expressed at different growth stages and in response to light, temperature, and drought. Only SmTIR1A expression was significantly induced by picloram treatment and different growth stages. TIR1/AFB expression is regulated by microRNAs (miRNAs) in other plant species, and we identified 6 or 29 miRNAs that potentially targeted the five TIR1/AFB genes on the basis of comparisons with S. lycopersicum and S. tuberosum miRNAs, respectively. Three-dimensional protein structure predictions revealed that all the TIR1/AFB proteins were very similar in structure, differing only in the numbers of alpha helices and in one angle linking an α helix and a β sheet. For measuring the function of TIR1/AFB genes in response to drought, SmAFB5 was selected, and knockdown by virus-induced gene silence (VIGS) 35S::SmAFB5 lines showed resistance to drought compared to controls. These analyses provide insight into the potential functions of TIR1/AFBs during growth and in response to stress; they highlight differences among the SmTIR1/AFBs that may be useful for eggplant breeding.

RiboDiffusion: tertiary structure-based RNA inverse folding with generative diffusion models.

RNA design shows growing applications in synthetic biology and therapeutics, driven by the crucial role of RNA in various biological processes. A fundamental challenge is to find functional RNA sequences that satisfy given structural constraints, known as the inverse folding problem. Computational approaches have emerged to address this problem based on secondary structures. However, designing RNA sequences directly from 3D structures is still challenging, due to the scarcity of data, the nonunique structure-sequence mapping, and the flexibility of RNA conformation. In this study, we propose RiboDiffusion, a generative diffusion model for RNA inverse folding that can learn the conditional distribution of RNA sequences given 3D backbone structures. Our model consists of a graph neural network-based structure module and a Transformer-based sequence module, which iteratively transforms random sequences into desired sequences. By tuning the sampling weight, our model allows for a trade-off between sequence recovery and diversity to explore more candidates. We split test sets based on RNA clustering with different cut-offs for sequence or structure similarity. Our model outperforms baselines in sequence recovery, with an average relative improvement of 11% for sequence similarity splits and 16% for structure similarity splits. Moreover, RiboDiffusion performs consistently well across various RNA length categories and RNA types. We also apply in silico folding to validate whether the generated sequences can fold into the given 3D RNA backbones. Our method could be a powerful tool for RNA design that explores the vast sequence space and finds novel solutions to 3D structural constraints. The source code is available at https://github.com/ml4bio/RiboDiffusion.