We present the results for CAPRI Round 54, the 5th joint CASP-CAPRI protein assembly prediction challenge. The Round offered 37 targets, including 14 homo-dimers, 3 homo-trimers, 13 hetero-dimers including 3 antibody-antigen complexes, and 7 large assemblies. On average ~70 CASP and CAPRI predictor groups, including more than 20 automatics servers, submitted models for each target. A total of 21941 models submitted by these groups and by 15 CAPRI scorer groups were evaluated using the CAPRI model quality measures and the DockQ score consolidating these measures. The prediction performance was quantified by a weighted score based on the number of models of acceptable quality or higher submitted by each group among their 5 best models. Results show substantial progress achieved across a significant fraction of the 60+ participating groups. High-quality models were produced for about 40% for the targets compared to 8% two years earlier, a remarkable improvement resulting from the wide use of the AlphaFold2 and AlphaFold-Multimer software. Creative use was made of the deep learning inference engines affording the sampling of a much larger number of models and enriching the multiple sequence alignments with sequences from various sources. Wide use was also made of the AlphaFold confidence metrics to rank models, permitting top performing groups to exceed the results of the public AlphaFold-Multimer version used as a yard stick. This notwithstanding, performance remained poor for complexes with antibodies and nanobodies, where evolutionary relationships between the binding partners are lacking, and for complexes featuring conformational flexibility, clearly indicating that the prediction of protein complexes remains a challenging problem.

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.

CASP assessments primarily rely on comparing predicted coordinates with experimental reference structures. However, errors in the reference structures can potentially reduce the accuracy of the assessment. This issue is particularly prominent in cryoEM-determined structures, and therefore, in the assessment of CASP15 cryoEM targets, we directly utilized density maps to evaluate the predictions. A method for ranking the quality of protein chain predictions based on rigid fitting to experimental density was found to correlate well with the CASP assessment scores. Overall, the evaluation against the density map indicated that the models are of high accuracy although local assessment of predicted side chains in a 1.52 Å resolution map showed that side-chains are sometimes poorly positioned. The top 136 predictions associated with 9 protein target reference structures were selected for refinement, in addition to the top 40 predictions for 11 RNA targets. To this end, we have developed an automated hierarchical refinement pipeline in cryoEM maps. For both proteins and RNA, the refinement of CASP15 predictions resulted in structures that are close to the reference target structure, including some regions with better fit to the density. This refinement was successful despite large conformational changes and secondary structure element movements often being required, suggesting that predictions from CASP-assessed methods could serve as a good starting point for building atomic models in cryoEM maps for both proteins and RNA. Loop modeling continued to pose a challenge for predictors with even short loops failing to be accurately modeled or refined at times. The lack of consensus amongst models suggests that modeling holds the potential for identifying more flexible regions within the structure.

The canonical function of glutamyl-tRNA synthetase (GluRS) is to glutamylate tRNA Glu. Yet, not all bacterial GluRSs glutamylate tRNA Glu; many glutamylate both tRNA Glu and tRNA Gln, while some glutamylate only tRNA Gln and not the cognate substrate tRNA Glu. Understanding the basis of this unique tRNA Glx-specificity is important. Mutational studies have hinted at hotspot residues, both on tRNA Glx and GluRS, that play crucial roles in tRNA Glx-specificity. But the underlying structural basis remains unexplored. Majority of biochemical studies related to tRNA Glx-specificity have been performed on GluRS from Escherichia coli and other proteobacterial species. However, since the early crystal structures of GluRS and tRNA Glu-bound GluRS were from non-proteobacterial species ( Thermus thermophilus), the proteobacterial biochemical data have often been interpreted in the context of non-proteobacterial GluRS structures. Marked differences between proteo- and non-proteobacterial GluRSs have been demonstrated and therefore it is important that tRNA Glx-specificity be understood vis-a-vis proteobacterial GluRS structures. Towards this goal we have solved the crystal structure of GluRS from E. coli. Using the solved structure and several other currently available proteo- and non-proteobacterial GluRS crystal structures, we have probed the structural basis of tRNA Glx-specificity of bacterial GluRSs. Specifically, our analysis suggests a unique role played by a tRNA Glx D-helix contacting loop of GluRS in modulation of tRNA Gln-specificity. While earlier studies had identified functional hotspots on tRNA Glx that controlled tRNA Glx-specificity of GluRS, this is the first report of complementary signatures of tRNA Glx-specificity in GluRS.

Protein domains are structural, functional, and evolutionary units. These domains bring out the diversity of functionality by means of interactions with other co-existing domains and provide stability. Hence, it is important to study intra-protein inter-domain interactions from the perspective of types of interactions. Domains within a chain could interact over short timeframes or permanently, rather like protein-protein interactions (PPIs). However, no systematic study has been carried out between two classes, namely permanent and transient domain-domain interactions (DDIs). In this work, we studied 264 two-domain proteins, belonging to either of these classes and their interfaces on the basis of several factors, such as interface area and details of interactions (number, strengths, and types of interactions). We also characterized them based on residue conservation at the interface, correlation of residue motions across domains, its involvement in repeat formation, and their involvement in particular molecular processes. Finally, we could analyse the interactions arising from domains in two-domain monomeric proteins, and we observed significant differences between these two classes of domain interactions and a few similarities. This study will help to obtain a better understanding of structure-function and folding principles of multi-domain proteins.

Proteins such as enzymes perform their function by predominant non-covalent bond interactions between transiently interacting units. There is an impact on the overall structural topology of the protein, albeit transient nature of such interactions, that enable proteins to deactivate or activate. This aspect of the alteration of the structural topology is studied by employing protein structural networks, which are node-edge representative models of protein structure, reported as a robust tool for capturing interactions between residues. Several methods have been optimised to collect meaningful, functionally relevant information by studying alteration of structural networks. In this article, different methods of comparing protein structural networks are employed, along with spectral decomposition of graphs to study the subtle impact of protein-protein interactions. A detailed analysis of the structural network of interacting partners is performed across a dataset of around 900 pairs of bound complexes and corresponding unbound protein structures. The variation in network parameters at, around and far away from the interface are analysed. Finally, we present interesting case studies, where an allosteric mechanism of structural impact is understood from communication-path detection methods. The results of this analysis are beneficial in understanding protein stability, for future engineering and docking studies.

Trimethylamine monooxygenase ( Tmm, EC-1.14.13.148) belongs to the family of flavin-containing monooxygenases (FMOs) that oxidize trimethylamine into trimethylamine-N-oxide (TMAO). Conventional methods for assaying Tmm are accurate over a narrow range of substrate/ product concentrations. Here we report a TMAO-specific enzymatic assay for Tmm using polyallylamine hydrochloride (PAHCl)-capped MnO 2 nanoparticles (PAHCl@MnO 2). We achieved TMAO specificity using iodoacetonitrile to remove interfering trimethylamine. The change in the concentration of TMAO is measured by observing the difference in the absorbance of 3,3´,5,5´-tetramethylbenzidine (TMB) at 652 nm. The assay is tolerant to several interfering metal ions and other compounds. This method is more reliable and easier than currently known methods. The limit of detection (LOD) and limit of quantitation (LOQ) are 1 µM and 10 µM, respectively, for direct TMAO measurement.

The results of tertiary structure assessment at CASP15 are reported. For the first time, recognising the outstanding performance of AlphaFold 2 (AF2) at CASP14, all single chain predictions were assessed together, irrespective of whether a template was available. At CASP15 there was no single stand-out group, with most of the best-scoring groups - led by PEZYFoldings, UM-TBM and Yang Server - employing AF2 in one way or another. Many top groups paid special attention to generating deep Multiple Sequence Alignments (MSAs) and testing variant MSAs, thereby allowing them to successfully address some of the hardest targets. Such difficult targets, as well as lacking templates, were typically proteins with few homologues: small size, high α-helical content and monomeric structure were other likely aggravating factors. Local divergence between prediction and target correlated with localisation at crystal lattice or chain interfaces, and with regions exhibiting high B-factor factors in crystal structure targets, but should not necessarily be considered as representing error in the prediction. However, analysis of exposed and buried side chain accuracy showed room for improvement even in the latter. Nevertheless, a majority of groups, including those applying methods similar to those used to generate major resources such as the AlphaFold Protein Structure Database and the ESM Metagenomic atlas, produced high quality predictions for most targets which are valuable for experimental structure determination, functional analysis and many other tasks across biology.

The core metabolic reactions of life drive electrons through a class of redox protein enzymes, the oxidoreductases. The energetics of electron flow is determined by the redox potentials of organic and inorganic cofactors as tuned by the protein environment. Understanding how protein structure affects oxidation-reduction energetics is crucial for studying metabolism, creating bioelectronic systems, and tracing the history of biological energy utilization on Earth. We constructed ProtReDox ([https://protein-redox-potential.web.app](https://protein-redox-potential.web.app)), a manually curated database of experimentally determined redox potentials. With over 500 measurements, we can begin to identify how proteins modulate oxidation-reduction energetics across the tree of life. By mapping redox potentials onto networks of oxidoreductase fold evolution, we can infer the evolution of electron transfer energetics over deep-time. ProtReDox is designed to include user-contributed submissions with the intention of making it a valuable resource for researchers in this field.

The prediction of protein-ligand complexes (PLC), using both experimental and predicted structures, is an active and important area of research, underscored by the inclusion of the Protein-Ligand Interaction category in the latest round of the Critical Assessment of Protein Structure Prediction experiment CASP15. The prediction task in CASP15 consisted of predicting both the 3-dimensional structure of the receptor protein as well as the position and conformation of the ligand. This paper addresses the challenges and proposed solutions for devising automated benchmarking techniques for PLC prediction. The reliability of experimentally solved PLC as ground truth reference structures is assessed using various validation criteria. Similarity of PLC to previously released complexes are employed to judge the novelty and difficulty of a PLC as a prediction target. We show that the commonly used PDBBind time-split test-set is inappropriate for comprehensive PLC evaluation. Finally, we introduce a fully automated pipeline that predicts PLC and evaluates the accuracy of the protein structure, ligand pose, and protein-ligand interactions.

Prediction categories in the Critical Assessment of Structure Prediction (CASP) experiments change with the need to address specific problems in structure modeling. In CASP15, four new prediction categories were introduced: RNA structure, ligand-protein complexes, accuracy of oligomeric structures and their interfaces, and ensembles of alternative conformations. This paper lists technical specifications for these categories and describes their integration in the CASP data management system.

This article reports and analyzes the results of protein complex model accuracy estimation by our methods (DeepUMQA3 and GraphGPSM) in the 15 th Critical Assessment of techniques for protein Structure Prediction (CASP15). The new deep learning-based multimeric complex model accuracy estimation methods are proposed based on the ensemble of three level features coupling with deep residual/graph neural networks. For the input multimeric complex model, we describe it from three levels: overall complex features, intra-monomer features, and inter-monomer features. We designed an overall ultrafast shape recognition (USR) to characterize the relationship between local residues and the overall complex topology, and an inter-monomer USR to characterize the relationship between the residues of one monomer and the topology of other monomers. On the 39 complex targets of CASP15, DeepUMQA3 (Group name: GuijunLab-RocketX) ranked first in the assessment of interface residue accuracy. The Pearson correlation coefficient (PCC) between the interface residues lDDT predicted by DeepUMQA3 and the real lDDT is 0.570, and DeepUMQA3 achieved the highest PCC on 29 out of 39 targets. GraphGPSM (Group name: GuijunLab-PAthreader) had a TM-score PCC>0.9 on 14 targets, showing a good ability to estimate the overall fold accuracy.

The human predictor team PEZYFoldings got third place with GDT-TS (First place with the Assessor’s formulae) in the single-domain category and tenth place in the multimer category in CASP15. In this paper, I describe the exact method used by PEZYFoldings in competitions. As AlphaFold2 and AlphaFold-Multimer, developed by DeepMind, are state-of-the-art structure prediction tools, it was assumed that enhancing the input and output of the tools was an effective strategy to obtain the highest accuracy for structure prediction. Therefore, I used additional tools and databases to collect evolutionarily related sequences and introduced a deep-learning-based model in the refinement step. In addition to these modifications, manual interventions were performed to address various tasks. Detailed analyses were performed after the competition to identify the main contributors to performance. Comparing the number of evolutionarily related sequences I used with those of the other teams that provided AlphaFold2’s baseline predictions revealed that an extensive sequence similarity search was one of the main contributors. The impact of the refinement model was minimal (p <0.05 for the TM score). In addition, I noticed that I had gained large Z-scores with the subunits of H1137, for which I performed manual domain parsing considering the interfaces between the subunits. This finding implies that the manual intervention contributed to my performance. The prediction performance was low when I could not identify the evolutionarily related sequences. T1130 is an example; however, other teams can model better structures. Based on the discussions from the CASP15 conference, the two teams that ranked higher than PEZYFoldings had some hits for T1130. This may be because T1130 is a eukaryotic protein, whereas the additional databases used were mainly from metagenomic sequences, which primarily consist of prokaryotic proteins. These results highlight the opportunities for improvement in 1) multimer prediction, 2) building larger and more diverse databases, and 3) developing tools to predict structures from primary sequences alone. In addition, transferring the manual intervention process to automation is a future concern.

We introduce a deep learning-based ligand pose scoring model called zPoseScore for predicting protein-ligand complexes in the 15th Critical Assessment of Protein Structure Prediction (CASP15). Our contributions are three-fold: firstly, we generate six training and evaluation datasets by employing advanced data augmentation and sampling methods. Secondly, we redesign the “zFormer” module, inspired by AlphaFold2’s Evoformer, to efficiently describe protein-ligand interactions. This module enables the extraction of protein-ligand paired features that lead to accurate predictions. Lastly, we develop the zPoseScore framework with zFormer for scoring and ranking ligand poses, allowing for atomic-level protein-ligand feature encoding and fusion to output refined ligand poses and ligand per-atom deviations. Our results demonstrate excellent performance on various testing datasets, achieving Pearson’s correlation R = 0.783 and 0.659 for ranking docking decoys generated based on experimental and predicted protein structures of CASF-2016 protein-ligand complexes. Additionally, we obtain an averaged lDDT = 0.558 of AIchemy_LIG2 in CASP15 for de novo protein-ligand complex structure predictions. Detailed analysis shows that accurate ligand binding site prediction and side-chain orientation are crucial for achieving better prediction performance. Our proposed model is one of the most accurate protein-ligand pose prediction models and could serve as a valuable tool in small molecule drug discovery.

AlphaFold has transformed structure prediction by enabling highly accurate predictions on par with experimentally determined structures. Still, for difficult cases, in particular, multimers, there is still room for improvement. Important for the success of AlphaFold is its ability to assess its own predictions. The basic idea for the Wallner group in CASP15 was to exploit the excellent ranking score in AlphaFold by massive sampling. To this end, we ran AlphaFold using six different settings, with and without templates, and with an increased number of recycles using both multimer v1 and v2 weights. In all cases, the dropout layers were enabled at inference to sample the uncertainty and increase the diversity of the generated models. A median of 4,810 models per target was generated and almost all (35/38) received a ranking_confidence >0.7. Compared to other groups in CASP15, Wallner obtained the highest sum of Z-scores based on the DockQ score, 40.8 compared to 26.3 for the second highest, much higher than -0.2 achieved by the AlphaFold baseline method, NBIS-AF2-multimer. The improvement over the baseline is substantial with the mean DockQ increasing from 0.43 to 0.56, with several targets showing a DockQ score increase by +0.6 units. Remarkable, considering Wallner and NBIS-AF2-multimer were using identical input data. The reason for the success can be attributed to the diversified sampling using dropout with different settings and, in particular, the use of multimer v1, which seems to be much more susceptible to sampling compared to v2. The method is available here: http://wallnerlab.org/AFsample/.

In CASP15 there was a greater emphasis on multimeric modelling than in previous experiments, with assembly structures nearly doubling in number (41 up from 22) since the previous round. CASP15 also included a new estimation of model accuracy (EMA) category in recognition of the importance of objective quality assessment for quaternary structure models. ModFOLDdock is a multimeric model quality assessment server developed by the McGuffin group at the University of Reading, which brings together a range of single-model, clustering and deep learning methods to form a consensus of approaches. For CASP15 three variants of ModFOLDdock were developed to optimise for the different facets of the quality estimation problem. The standard ModFOLDdock variant produced predicted scores optimised for positive linear correlations with the observed scores. The ModFOLDdockR variant produced predicted scores optimised for ranking, i.e., the top-ranked models have highest accuracy. In addition, the ModFOLDdockS variant used a quasi-single model approach to score each model on an individual basis. The scores from all three variants achieved strongly positive Pearson correlation coefficients with the CASP observed scores (oligo-lDDT) in excess of 0.70, which were maintained across both homomeric and heteromeric model populations. In addition, at least one of the ModFOLDdock variants was consistently ranked in the top two methods across all three EMA categories. Specifically, for overall global fold prediction accuracy, ModFOLDdock placed second and ModFOLDdockR placed third; for overall interface quality prediction accuracy ModFOLDdockR, ModFOLDdock and ModFOLDdockS were placed above all other predictor methods, and ModFOLDdockR and ModFOLDdockS were placed second and third respectively for individual residue confidence scores. The ModFOLDdock server is available at: [https://www.reading.ac.uk/bioinf/ModFOLDdock/](https://www.reading.ac.uk/bioinf/ModFOLDdock/). ModFOLDdock is also available as part of the MultiFOLD docker package: [https://hub.docker.com/r/mcguffin/multifold](https://hub.docker.com/r/mcguffin/multifold)

Leucine and Isoleucine are two amino acids that differ only by the positioning of one methyl group. This small difference has however important consequences in α-helices, as the β-branching of Ile results in helix destabilization. We set out to investigate whether there are general trends for the occurrences of Leu and Ile residues in structures and sequences of class A GPCRs (G protein-coupled receptors). GPCRs are integral membrane proteins in which α-helices span the plasma membrane seven times and which play a crucial role in signal transmission into the cell. We found that Leu side chains are generally present in less densely packed regions and are more protein-surface exposed than Ile side chains. We explored whether this difference might be attributed to different functions of the two amino acids and tested if Leu adjusts the hydrophobicity of the transmembrane domain based on the Wimley-White whole-residue hydrophobicity scales. In class A GPCRs, Leu decreases the variation in hydropathy between receptors and Leu content correlates positively with hydropathy calculated without Leu. Both measures indicate that hydropathy is tuned by Leu. To test this idea further, we generated protein sequences with random amino acid compositions using a simple numerical model, in which hydropathy was tuned by adjusting the number of Leu residues. The model was able to replicate the observations made with class A GPCR sequences. We speculate that Leu tunes the hydropathy of the transmembrane domain of class A GPCRs to facilitate correct insertion into membranes and/or for stability within them.

Estimating the accuracy of quaternary structural models of protein complexes and assemblies (EMA) is important for predicting quaternary structures and applying them to studying protein function and interaction. The pairwise similarity between structural models is proven useful for estimating the quality of protein tertiary structural models, but it has been rarely applied to predicting the quality of quaternary structural models. Moreover, the pairwise similarity approach often fails when many structural models are of low quality and similar to each other. To address the gap, we developed a hybrid method (MULTICOM_qa) combining a pairwise similarity score (PSS) and an interface contact probability score (ICPS) based on the deep learning inter-chain contact prediction for estimating protein complex model accuracy. It blindly participated in the 15th Critical Assessment of Techniques for Protein Structure Prediction (CASP15) in 2022 and ranked first out of 24 predictors in estimating the global accuracy of assembly models. The average per-target correlation coefficient between the model quality scores predicted by MULTICOM_qa and the true quality scores of the models of CASP15 assembly targets is 0.66. The average per-target ranking loss in using the predicted quality scores to rank the models is 0.14. It was able to select good models for most targets. Moreover, several key factors (i.e., target difficulty, model sampling difficulty, skewness of model quality, and similarity between good/bad models) for EMA are identified and analayzed. The results demonstrate that combining the multi-model method (PSS) with the complementary single-model method (ICPS) is a promising approach to EMA.