Comparative analysis of RNA 3D structure prediction methods: towards enhanced modeling of RNA-ligand interactions.

Accurate RNA structure models are crucial for designing small molecule ligands that modulate their functions. This study assesses six standalone RNA 3D structure prediction methods-DeepFoldRNA, RhoFold, BRiQ, FARFAR2, SimRNA and Vfold2, excluding web-based tools due to intellectual property concerns. We focus on reproducing the RNA structure existing in RNA-small molecule complexes, particularly on the ability to model ligand binding sites. Using a comprehensive set of RNA structures from the PDB, which includes diverse structural elements, we found that machine learning (ML)-based methods effectively predict global RNA folds but are less accurate with local interactions. Conversely, non-ML-based methods demonstrate higher precision in modeling intramolecular interactions, particularly with secondary structure restraints. Importantly, ligand-binding site accuracy can remain sufficiently high for practical use, even if the overall model quality is not optimal. With the recent release of AlphaFold 3, we included this advanced method in our tests. Benchmark subsets containing new structures, not used in the training of the tested ML methods, show that AlphaFold 3's performance was comparable to other ML-based methods, albeit with some challenges in accurately modeling ligand binding sites. This study underscores the importance of enhancing binding site prediction accuracy and the challenges in modeling RNA-ligand interactions accurately.

The ubiquitin-specific protease 21 is critical for cancer cell mitochondrial function and regulates proliferation and migration.

Ubiquitin-Specific Peptidases (USPs) are the main members of deubiquitinases (DUBs) that catalyze removing ubiquitin chains from target proteins, thereby modulating their half-life and function. Enzymatic activity of USP21 regulates protein degradation which is critical for maintaining cell homeostasis. USP21 determines the stability of oncogenic proteins and therefore is implicated in carcinogenesis. In this study, we investigated the effect of USP21 deletion on cancer cell metabolism. Transcriptomic and proteomic analysis of USP21 knockout HAP-1 cells revealed that endogenous USP21 is critical for the expression of genes and proteins involved in mitochondrial function. Additionally, we have found that deletion of USP21 reduced STAT3 activation and STAT3-dependent gene and protein expression in cancer cells. Genetic deletion of USP21 impaired mitochondrial respiration and disturbed ATP production. This resulted in cellular consequences such as inhibition of cell proliferation and migration. Presented results provide new insights into the biology of USP21, suggesting novel mechanisms for controlling STAT3 activity and mitochondrial function in tumor cells. Taken together, our findings indicate that targeting USP21 dysregulates the energy status of cancer cells offering new perspectives for anti-cancer therapy.

QChromosomeVisualizer: A new tool for 3D visualization of long simulations of polymer-like chromosome models.

Recent years have brought us great wealth of new types of experimental data on different aspects of chromatin state, from chromosome conformation assays, through super-resolution microscopic imaging to epigenetic modifications and lamina interaction assays. This rapid increase in data availability have motivated many novel approaches to 3D modeling of chromosomes, their conformations and dynamic behavior. Even though there are many tools already developed for molecular visualization in the field of structural bioinformatics, they are usually optimized for visualization of smaller molecules (like proteins) and much shorter trajectories. We have developed a novel approach to visualization of long trajectories of large polymers, typical in the field of chromatin modeling. Our software, called QChromosomeVisualizer (QCV), allows for quick visualization of long simulations containing thousands or even millions of frames and generating good looking still images and animations including spherical 360 videos that can be viewed in VR headsets. We believe that this kind of tools will be helpful for the broader community of researchers interested in modeling by allowing them to create new and clearer ways to communicate their results.

Endothelial cell differentiation is encompassed by changes in long range interactions between inactive chromatin regions.

Endothelial cells (ECs) differentiate from mesodermal progenitors during vasculogenesis. By comparing changes in chromatin interactions between human umbilical vein ECs, embryonic stem cells and mesendoderm cells, we identified regions exhibiting EC-specific compartmentalization and changes in the degree of connectivity within topologically associated domains (TADs). These regions were characterized by EC-specific transcription, binding of lineage-determining transcription factors and cohesin. In addition, we identified 1200 EC-specific long-range interactions (LRIs) between TADs. Most of the LRIs were connected between regions enriched for H3K9me3 involving pericentromeric regions, suggesting their involvement in establishing compartmentalization of heterochromatin during differentiation. Second, we provide evidence that EC-specific LRIs correlate with changes in the hierarchy of chromatin aggregation. Despite these rearrangements, the majority of chromatin domains fall within a pre-established hierarchy conserved throughout differentiation. Finally, we investigated the effect of hypoxia on chromatin organization. Although hypoxia altered the expression of hundreds of genes, minimal effect on chromatin organization was seen. Nevertheless, 70% of hypoxia-inducible genes situated within a TAD bound by HIF1α suggesting that transcriptional responses to hypoxia largely depend on pre-existing chromatin organization. Collectively our results show that large structural rearrangements establish chromatin architecture required for functional endothelium and this architecture remains largely unchanged in response to hypoxia.

NPDock: a web server for protein-nucleic acid docking.

Protein-RNA and protein-DNA interactions play fundamental roles in many biological processes. A detailed understanding of these interactions requires knowledge about protein-nucleic acid complex structures. Because the experimental determination of these complexes is time-consuming and perhaps futile in some instances, we have focused on computational docking methods starting from the separate structures. Docking methods are widely employed to study protein-protein interactions; however, only a few methods have been made available to model protein-nucleic acid complexes. Here, we describe NPDock (Nucleic acid-Protein Docking); a novel web server for predicting complexes of protein-nucleic acid structures which implements a computational workflow that includes docking, scoring of poses, clustering of the best-scored models and refinement of the most promising solutions. The NPDock server provides a user-friendly interface and 3D visualization of the results. The smallest set of input data consists of a protein structure and a DNA or RNA structure in PDB format. Advanced options are available to control specific details of the docking process and obtain intermediate results. The web server is available at http://genesilico.pl/NPDock.

Computational modeling of protein-RNA complex structures.

Protein-RNA interactions play fundamental roles in many biological processes, such as regulation of gene expression, RNA splicing, and protein synthesis. The understanding of these processes improves as new structures of protein-RNA complexes are solved and the molecular details of interactions analyzed. However, experimental determination of protein-RNA complex structures by high-resolution methods is tedious and difficult. Therefore, studies on protein-RNA recognition and complex formation present major technical challenges for macromolecular structural biology. Alternatively, protein-RNA interactions can be predicted by computational methods. Although less accurate than experimental measurements, theoretical models of macromolecular structures can be sufficiently accurate to prompt functional hypotheses and guide e.g. identification of important amino acid or nucleotide residues. In this article we present an overview of strategies and methods for computational modeling of protein-RNA complexes, including software developed in our laboratory, and illustrate it with practical examples of structural predictions.

RNA-Puzzles: a CASP-like evaluation of RNA three-dimensional structure prediction.

We report the results of a first, collective, blind experiment in RNA three-dimensional (3D) structure prediction, encompassing three prediction puzzles. The goals are to assess the leading edge of RNA structure prediction techniques; compare existing methods and tools; and evaluate their relative strengths, weaknesses, and limitations in terms of sequence length and structural complexity. The results should give potential users insight into the suitability of available methods for different applications and facilitate efforts in the RNA structure prediction community in ongoing efforts to improve prediction tools. We also report the creation of an automated evaluation pipeline to facilitate the analysis of future RNA structure prediction exercises.

Crystal structures of the tRNA:m2G6 methyltransferase Trm14/TrmN from two domains of life.

Methyltransferases (MTases) form a major class of tRNA-modifying enzymes needed for the proper functioning of tRNA. Recently, RNA MTases from the TrmN/Trm14 family that are present in Archaea, Bacteria and Eukaryota have been shown to specifically modify tRNA(Phe) at guanosine 6 in the tRNA acceptor stem. Here, we report the first X-ray crystal structures of the tRNA m(2)G6 (N(2)-methylguanosine) MTase (TTC)TrmN from Thermus thermophilus and its ortholog (Pf)Trm14 from Pyrococcus furiosus. Structures of (Pf)Trm14 were solved in complex with the methyl donor S-adenosyl-l-methionine (SAM or AdoMet), as well as the reaction product S-adenosyl-homocysteine (SAH or AdoHcy) and the inhibitor sinefungin. (TTC)TrmN and (Pf)Trm14 consist of an N-terminal THUMP domain fused to a catalytic Rossmann-fold MTase (RFM) domain. These results represent the first crystallographic structure analysis of proteins containing both THUMP and RFM domain, and hence provide further insight in the contribution of the THUMP domain in tRNA recognition and catalysis. Electrostatics and conservation calculations suggest a main tRNA binding surface in a groove between the THUMP domain and the MTase domain. This is further supported by a docking model of TrmN in complex with tRNA(Phe) of T. thermophilus and via site-directed mutagenesis.

Sequence-specific cleavage of the RNA strand in DNA-RNA hybrids by the fusion of ribonuclease H with a zinc finger.

Ribonucleases (RNases) are valuable tools applied in the analysis of RNA sequence, structure and function. Their substrate specificity is limited to recognition of single bases or distinct secondary structures in the substrate. Currently, there are no RNases available for purely sequence-dependent fragmentation of RNA. Here, we report the development of a new enzyme that cleaves the RNA strand in DNA-RNA hybrids 5 nt from a nonanucleotide recognition sequence. The enzyme was constructed by fusing two functionally independent domains, a RNase HI, that hydrolyzes RNA in DNA-RNA hybrids in processive and sequence-independent manner, and a zinc finger that recognizes a sequence in DNA-RNA hybrids. The optimization of the fusion enzyme's specificity was guided by a structural model of the protein-substrate complex and involved a number of steps, including site-directed mutagenesis of the RNase moiety and optimization of the interdomain linker length. Methods for engineering zinc finger domains with new sequence specificities are readily available, making it feasible to acquire a library of RNases that recognize and cleave a variety of sequences, much like the commercially available assortment of restriction enzymes. Potentially, zinc finger-RNase HI fusions may, in addition to in vitro applications, be used in vivo for targeted RNA degradation.

Effective modeling of the chromatin structure by coarse-grained methods.

The interphase chromatin structure is extremely complex, precise and dynamic. Experimental methods can only show the frequency of interaction of the various parts of the chromatin. Therefore, it is extremely important to develop theoretical methods to predict the chromatin structure. In this publication, we implemented an extended version of the SBS model described by Barbieri et al. and created the ChroMC program that is easy to use and freely available (https://github.com/regulomics/chroMC) to other users. We also describe the necessary factors for the effective modeling of the chromatin structure in Drosophila melanogaster. We compared results of chromatin structure predictions using two methods: Monte Carlo and Molecular Dynamic. Our simulations suggest that incorporating black, non-reactive chromatin is necessary for successful prediction of chromatin structure, while the loop extrusion model with a long range attraction potential or Lennard-Jones (with local attraction force) as well as using Hi-C data as input are not essential for the basic structure reconstruction. We also proposed a new way to calculate the similarity of the properties of contact maps including the calculation of local similarity.Communicated by Ramaswamy H. Sarma.

Computational methods for prediction of protein-RNA interactions.

Understanding the molecular mechanism of protein-RNA recognition and complex formation is a major challenge in structural biology. Unfortunately, the experimental determination of protein-RNA complexes by X-ray crystallography and nuclear magnetic resonance spectroscopy (NMR) is tedious and difficult. Alternatively, protein-RNA interactions can be predicted by computational methods. Although less accurate than experimental observations, computational predictions can be sufficiently accurate to prompt functional hypotheses and guide experiments, e.g. to identify individual amino acid or nucleotide residues. In this article we review 10 methods for predicting protein-RNA interactions, seven of which predict RNA-binding sites from protein sequences and three from structures. We also developed a meta-predictor that uses the output of top three sequence-based primary predictors to calculate a consensus prediction, which outperforms all the primary predictors. In order to fully cover the software for predicting protein-RNA interactions, we also describe five methods for protein-RNA docking. The article highlights the strengths and shortcomings of existing methods for the prediction of protein-RNA interactions and provides suggestions for their further development.

DARS-RNP and QUASI-RNP: new statistical potentials for protein-RNA docking.

BACKGROUND: Protein-RNA interactions play fundamental roles in many biological processes. Understanding the molecular mechanism of protein-RNA recognition and formation of protein-RNA complexes is a major challenge in structural biology. Unfortunately, the experimental determination of protein-RNA complexes is tedious and difficult, both by X-ray crystallography and NMR. For many interacting proteins and RNAs the individual structures are available, enabling computational prediction of complex structures by computational docking. However, methods for protein-RNA docking remain scarce, in particular in comparison to the numerous methods for protein-protein docking. RESULTS: We developed two medium-resolution, knowledge-based potentials for scoring protein-RNA models obtained by docking: the quasi-chemical potential (QUASI-RNP) and the Decoys As the Reference State potential (DARS-RNP). Both potentials use a coarse-grained representation for both RNA and protein molecules and are capable of dealing with RNA structures with posttranscriptionally modified residues. We compared the discriminative power of DARS-RNP and QUASI-RNP for selecting rigid-body docking poses with the potentials previously developed by the Varani and Fernandez groups. CONCLUSIONS: In both bound and unbound docking tests, DARS-RNP showed the highest ability to identify native-like structures. Python implementations of DARS-RNP and QUASI-RNP are freely available for download at http://iimcb.genesilico.pl/RNP/

FILTREST3D: discrimination of structural models using restraints from experimental data.

SUMMARY: Automatic methods for macromolecular structure prediction (fold recognition, de novo folding and docking programs) produce large sets of alternative models. These large model sets often include many native-like structures, which are often scored as false positives. Such native-like models can be more easily identified based on data from experimental analyses used as structural restraints (e.g. identification of nearby residues by cross-linking, chemical modification, site-directed mutagenesis, deuterium exchange coupled with mass spectrometry, etc.). We present a simple server for scoring and ranking of models according to their agreement with user-defined restraints. AVAILABILITY: FILTREST3D is freely available for users as a web server and standalone software at: http://filtrest3d.genesilico.pl/ CONTACT: iamb@genesilico.pl SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The structure of M.EcoKI Type I DNA methyltransferase with a DNA mimic antirestriction protein.

Type-I DNA restriction-modification (R/M) systems are important agents in limiting the transmission of mobile genetic elements responsible for spreading bacterial resistance to antibiotics. EcoKI, a Type I R/M enzyme from Escherichia coli, acts by methylation- and sequence-specific recognition, leading to either methylation of DNA or translocation and cutting at a random site, often hundreds of base pairs away. Consisting of one specificity subunit, two modification subunits, and two DNA translocase/endonuclease subunits, EcoKI is inhibited by the T7 phage antirestriction protein ocr, a DNA mimic. We present a 3D density map generated by negative-stain electron microscopy and single particle analysis of the central core of the restriction complex, the M.EcoKI M(2)S(1) methyltransferase, bound to ocr. We also present complete atomic models of M.EcoKI in complex with ocr and its cognate DNA giving a clear picture of the overall clamp-like operation of the enzyme. The model is consistent with a large body of experimental data on EcoKI published over 40 years.

HiCEnterprise: identifying long range chromosomal contacts in Hi-C data.

MOTIVATION: Computational analysis of chromosomal contact data is currently gaining popularity with the rapid advance in experimental techniques providing access to a growing body of data. An important problem in this area is the identification of long range contacts between distinct chromatin regions. Such loops were shown to exist at different scales, either mediating relatively short range interactions between enhancers and promoters or providing interactions between much larger, distant chromosome domains. A proper statistical analysis as well as availability to a wide research community are crucial in a tool for this task. RESULTS: We present HiCEnterprise, a first freely available software tool for identification of long range chromatin contacts not only between small regions, but also between chromosomal domains. It implements four different statistical tests for identification of significant contacts for user defined regions or domains as well as necessary functions for input, output and visualization of chromosome contacts. AVAILABILITY: The software and the corresponding documentation are available at: github.com/regulomics/HiCEnterprise. SUPPLEMENTARY INFORMATION: Supplemental data are available in the online version of the article and at the website regulomics.mimuw.edu.pl/wp/hicenterprise.

PROTMAP2D: visualization, comparison and analysis of 2D maps of protein structure.

MOTIVATION: Protein structure comparison is a fundamental problem in structural biology and bioinformatics. Two-dimensional maps of distances between residues in the structure contain sufficient information to restore the 3D representation, while maps of contacts reveal characteristic patterns of interactions between secondary and super-secondary structures and are very attractive for visual analysis. The overlap of 2D maps of two structures can be easily calculated, providing a sensitive measure of protein structure similarity. PROTMAP2D is a software tool for calculation of contact and distance maps based on user-defined criteria, quantitative comparison of pairs or series of contact maps (e.g. alternative models of the same protein, model versus native structure, different trajectories from molecular dynamics simulations, etc.) and visualization of the results. AVAILABILITY: PROTMAP2D for Windows / Linux / MacOSX is freely available for academic users from http://genesilico.pl/protmap2d.htm

THUMP from archaeal tRNA:m22G10 methyltransferase, a genuine autonomously folding domain.

The tRNA:m2(2)G10 methyltransferase of Pyrococus abyssi (PAB1283, a member of COG1041) catalyzes the N2,N2-dimethylation of guanosine at position 10 in tRNA. Boundaries of its THUMP (THioUridine synthases, RNA Methyltransferases and Pseudo-uridine synthases)--containing N-terminal domain [1-152] and C-terminal catalytic domain [157-329] were assessed by trypsin limited proteolysis. An inter-domain flexible region of at least six residues was revealed. The N-terminal domain was then produced as a standalone protein (THUMPalpha) and further characterized. This autonomously folded unit exhibits very low affinity for tRNA. Using protein fold-recognition (FR) methods, we identified the similarity between THUMPalpha and a putative RNA-recognition module observed in the crystal structure of another THUMP-containing protein (ThiI thiolase of Bacillus anthracis). A comparative model of THUMPalpha structure was generated, which fulfills experimentally defined restraints, i.e. chemical modification of surface exposed residues assessed by mass spectrometry, and identification of an intramolecular disulfide bridge. A model of the whole PAB1283 enzyme docked onto its tRNA(Asp) substrate suggests that the THUMP module specifically takes support on the co-axially stacked helices of T-arm and acceptor stem of tRNA and, together with the catalytic domain, screw-clamp structured tRNA. We propose that this mode of interactions may be common to other THUMP-containing enzymes that specifically modify nucleotides in the 3D-core of tRNA.

Novel inhibitors of the rRNA ErmC' methyltransferase to block resistance to macrolides, lincosamides, streptogramine B antibiotics.

In erythromycin-resistant bacteria, the N6 position of A2058 in 23S rRNA is mono- or dimethylated by Erm family methyltransferases. This modification results in cross-resistance to macrolides, lincosamides and streptogramin B. Most inhibitors of Erm methyltransferases developed up-to-date target the cofactor-binding pocket, resulting in a lack of selectivity whereas inhibitors that bind the substrate-binding pocket demonstrate low in vitro activity. In this study, a molecular docking approach followed by biochemical screening was applied to search for inhibitors targeting both cofactor- and substrate-binding pockets of ErmC' methyltransferase. Based on the results of the molecular docking-based virtual screening of the clean-leads subset of the ZINC database, 29 compounds were chosen for experimental verification. Among them inhibitor 28 (ZINC code 32747906), with an IC50 of 100 µM, decreased the minimal inhibitory concentration of erythromycin in the Escherichia coli strain overexpressing ErmC'. Docking analysis of 28 to the ErmC' structure and the competitive ligand binding assay revealed a non-competitive model of inhibition. Inhibitor 28 served as a template for similarity-based virtual screening, which resulted in the identification of two derivatives 3s (ZINC code 62022572) and 4s (ZINC code 49032257) with an IC50 of 116 µM and 110 µM, respectively. Our results provide a basis for the development of inhibitors against the Erm-family of enzymes.

Modeling of Protein-RNA Complex Structures Using Computational Docking Methods.

A significant part of biology involves the formation of RNA-protein complexes. X-ray crystallography has added a few solved RNA-protein complexes to the repertoire; however, it remains challenging to capture these complexes and often only the unbound structures are available. This has inspired a growing interest in finding ways to predict these RNA-protein complexes. In this study, we show ways to approach this problem by computational docking methods, either with a fully automated NPDock server or with a workflow of methods for generation of many alternative structures followed by selection of the most likely solution. We show that by introducing experimental information, the structure of the bound complex is rendered far more likely to be within reach. This study is meant to help the user of docking software understand how to grapple with a typical realistic problem in RNA-protein docking, understand what to expect in the way of difficulties, and recognize the current limitations.

Predicting atomic details of the unfolding pathway for YibK, a knotted protein from the SPOUT superfamily.

Several protein structures have been reported to contain intricate knots of the polypeptide backbone but the mechanism of the (un)folding process of knotted proteins remains unknown. The members of the SPOUT superfamily of RNA methyltransferases are some of the most intensely studied systems for investigation of the knot formation and function. YibK (whose biochemical function remains unknown) is the representative protein of the SPOUT superfamily. This protein exhibits a deep trefoil knot at the C-terminus. We conducted an extensive computational analysis of the unfolding process for the monomeric form of YibK. In order to predict the (un)folding pathway of YibK, we have calculated the order of secondary structure disassembly using UNFOLD, and performed thermal unfolding simulations using classical Molecular Dynamics (MD), as well as simulations employing reduced representation of the peptide chain using either MD with the UNRES method or the Monte Carlo (MC) unfolding with the REFINER method. Results obtained from all methods used in this work are in qualitative agreement. We found that YibK unfolds through four intermediate states. The trefoil knot in YibK disappears at the end of the unfolding process, long after the protein loses its native topology. We observed that the C- terminus leaves the knotting loop folded into a hairpin-like structure, in agreement with the results of coarse-grained simulation reported earlier. We propose that the folding pathway of YibK corresponds to the reversed sequence of events observed in the unfolding pathway elucidated in this study. Thus, we predict that the knot formation is the slowest part of the YibK folding process.