Despite recent progress, the Estimation of Model Accuracy (EMA) for protein complexes remains less advanced compared to that for protein monomers. A key challenge lies in effectively integrating both interface-specific and global structural information to accurately assess the quality of protein complexes. Here, we introduce ContrastQA, the first EMA framework for protein complexes that incorporates the proposed label-guided graph contrastive learning based on interface quality. By integrating a geometric graph neural network to model global structural features, ContrastQA effectively captures both local (interface-level) and global (structure-level) information for accurate model quality estimation. ContrastQA achieved ranking losses of 0.123 and 0.116 on the TMscore and GDT-TS metrics on the CASP16 dataset, which are 0.015 (10.9%) and 0.012 (8.7%) lower than the second-best EMA method with ranking losses of 0.138 and 0.128. Our study demonstrates the strong effectiveness of the label-guided graph contrastive learning module, particularly in selecting high-quality models. These findings suggest that our graph contrastive learning framework serves as a valuable pre-training strategy for learning protein structure representations.

Background:The Critical Assessment of Functional Annotation (CAFA) is a community effort held to understand the field of computational protein function prediction. Every three years, since 2010, the organizers initiate an experiment to collect function predictions on a large set of proteins and then evaluate the performance of predicting methods on a subset of proteins that have accumulated experimental annotations between the submission deadline and the evaluation time. CAFA provides an independent and rigorous assessment of the current state of the art, thus leveling the playing field, highlighting successes, revealing bottlenecks, and offering a forum for the exchange of ideas in protein science. Here, we report the results of the fourth CAFA experiment (CAFA4). Results:CAFA4 featured the participation of 148 methods from 70 research groups on a total of 46,205 unique proteins over a 5-year annotation accumulation phase, the longest in any CAFA. In a comparison across CAFA2-CAFA4 methods, the prediction of Gene Ontology (GO) terms has clearly improved across all three GO aspects and traditional evaluation settings. While not achieving the first rank, several CAFA2 and CAFA3 methods featured in the top ten methods in many evaluations, suggesting that earlier methods still hold relevance. The performance is weaker in the newly introduced "partial knowledge" evaluation category (proteins with experimental annotations before submission deadline that gained additional annotations in the same GO aspect during the annotation accumulation phase), highlighting the need for a new class of methods. The rankings of the methods were stable over the years in traditional evaluation settings, but less so in the new partial knowledge evaluation. Overall, the field continues to progress with some influx of new participants. Sustained efforts will be necessary to substantially advance it.

The goal of protein structure refinement is to enhance the precision of predicted protein models, particularly at the residue level of the local structure. Existing refinement approaches primarily rely on physics, whereas molecular simulation methods are resource-intensive and time-consuming. In this study, we employ deep learning methods to extract structural constraints from protein structure residues to assist in protein structure refinement. We introduce a novel method, AnglesRefine, which focuses on a protein's secondary structure and employs transformer to refine various protein structure angles (psi, phi, omega, CA_C_N_angle, C_N_CA_angle, N_CA_C_angle), ultimately generating a superior protein model based on the refined angles. We evaluate our approach against other cutting-edge methods using the CASP11-14 and CASP15 datasets. Experimental outcomes indicate that our method generally surpasses other techniques on the CASP11-14 test dataset, while performing comparably or marginally better on the CASP15 test dataset. Our method consistently demonstrates the least likelihood of model quality degradation, e.g., the degradation percentage of our method is less than 10%, while other methods are about 50%. Furthermore, as our approach eliminates the need for conformational search and sampling, it significantly reduces computational time compared to existing refinement methods.

Influence Maximization (IM) has been extensively studied in network science, which attempts to find a subset of users to maximize the influence spread. A new variant of IM, Fair Influence Maximization (FIM), which primarily enhances the fair propagation of information, attracts increasing attention in academic. However, existing algorithms for FIM suffer from a trade-off between fairness and running time. Since it is a tough task to ensure that users are fairly influenced in terms of sensitive attributes, such as race or gender, while maintaining a high influence spread. To tackle this problem, in this paper, we propose an effective and efficient Community-based Evolutionary Algorithm for FIM (named CEA-FIM). In CEA-FIM, a community-based node selection strategy is proposed to identify potential nodes, which not only considers the size of the community but also the attributes of the nodes in the community. Subsequently, we design an evolutionary algorithm based on the proposed node selection strategy to hasten the search for the optimal solution, including the novel initialization, crossover and mutation strategies. We validate the proposed algorithm CEA-FIM by performing experiments on real-world and synthetic networks. The experimental results show that the proposed CEA-FIM achieves a better balance between effectiveness and efficiency, compared to the state-of-the-art baseline algorithms.

MOTIVATION:With the advancement of deep learning, researchers have increasingly proposed computational methods based on deep learning techniques to predict protein function. However, many of these methods treat protein function prediction as a multi-label classification problem, often overlooking the long-tail distribution of functional labels (i.e., Gene Ontology Terms) in datasets. To address this issue, we propose the GOBoost method, which incorporates the proposed long-tail optimization ensemble strategy. Besides, GOBoost introduces the proposed global-local label graph module and multi-granularity focal loss function to enhance long-tail functional information, mitigate the long-tail phenomenon, and improve overall prediction accuracy. RESULTS:We evaluate GOBoost and other state-of-the-art (SOTA) protein function prediction methods on the PDB and AF2 datasets. The GOBoost outperformed SOTA methods on all evaluation metrics for both datasets. Notably, in the AUPR evaluation on the PDB test set, GOBoost improved by 10.71%, 35.91%, and 22.71% compared to the SOTA HEAL method in the MF, BP, and CC functions. The experimental results show the necessity and superiority of designing models from the label long-tail distribution perspective. AVAILABILITY AND IMPLEMENTATION:The source code of GOBoost is available at https://github.com/Cao-Labs/GOBoost.

In cryo-electron microscopy (cryo-EM) structural analysis, accurately predicting molecular conformations from image datasets is crucial. Despite advancements in deep learning, current solutions often produce suboptimal volume reconstructions that require additional processing for research use. To address this limitation, we introduce GraphConv, an innovative encoder model that embeds particle images into a latent space, replacing the conventional encoder used by CryoDRGN. Our approach employs a Graph Neural Network (GNN) architecture with multiple GraphConv and convolutional layers to capture richer information from particle images and accurately reconstruct 3D volumes. Rigorous testing on two authentic datasets and three simulated datasets demonstrates our model's efficacy, with improvements in reconstruction quality and resolution enhancements of up to 20 % compared to CryoDRGN. By leveraging the power of GNNs, our methodology significantly advances the fidelity and accuracy of cryo-EM volume reconstructions, contributing to the refinement of structural analysis techniques.

Proteins are fundamental to life, as they support vital processes in the body such as muscle development, cell growth, tissue repair, and immune defense. However, their complex structures and diverse functions make them challenging to fully understand. While recent advances enable efficient and accurate protein structure prediction, the challenge of predicting protein function remains. Although promising, current prediction methods suffer from slow performance, high computational demands, and struggle with handling highly specific proteins. Due to a rapid expansion of protein sequence databases, a computational method for predicting function directly from sequence is critical. Our solution to this ongoing challenge is Proteinext, an innovative method for protein function prediction that leverages advanced sequence representations and natural language processing (NLP) techniques. Proteinext leverages Meta's 15B-parameter evolutionary scale modeling to generate protein sequence embeddings, which are refined using a fine-tuned BigBird transformer-based NLP model. This combination results in a powerful model and method that significantly improves protein function prediction. The model was trained on 372,683 protein sequences from a combined dataset of Gene Ontology and Universal Protein Knowledgebase annotations. Proteinext represents a major step toward comprehensively understanding and predicting protein functions, achieving an impressive Fmax score of 0.74 and Smin score of 0.39. This work underscores the potential of combining computational biology with NLP to address critical challenges in proteomics. Proteinext is available at https://github.com/Cao-Labs/AlphaAnalyzer. Received: 18 March 2025 | Revised: 30 September 2025 | Accepted: 18 November 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in GitHub at https://github.com/Cao-Labs/AlphaAnalyzer. Author Contribution Statement Hailey Ledenko: Conceptualization, Methodology, Software, Writing – original draft, Writing – review & editing, Visualization. Luke Coleman: Conceptualization, Methodology, Software, Writing – original draft, Writing – review & editing. G. Alvarado: Writing – original draft, Writing – review & editing, Visualization. Tyler Stratton: Writing – original draft, Writing – review & editing. Boen Liu: Writing – original draft, Writing – review & editing. Jie Hou: Writing – original draft, Writing – review & editing. Dong Si: Writing – original draft, Writing – review & editing. Lei Zhang: Writing – original draft, Writing – review & editing. Rui Ding: Writing – original draft, Writing – review & editing. Yang Wang: Writing – original draft, Writing – review & editing. Renzhi Cao: Conceptualization, Writing – original draft, Writing – review & editing, Supervision, Project administration.

This study introduces ISBNet-Pocket, a novel method for ligand binding site prediction (LBSP) that utilizes 3D point cloud instance segmentation to address the limitations of clustering-based postprocessing in many LBSP models. ISBNet-Pocket represents proteins as 3D point clouds, enabling precise identification and separation of binding pockets without relying on clustering algorithms. This approach improves prediction accuracy, particularly for overlapping or closely located binding pockets, by directly localizing the ligand binding sites and their comprising atoms. Furthermore, a hybrid clustering strategy was tested to accommodate spatially proximate pockets, revealing both metric improvements and the inherent unreliability of clustering as a post-processing step. ISBNet-Pocket demonstrates performance consistent with state-of-the-art LBSP techniques while eliminating the need for clustering. Future research will refine the method and explore its application in molecular docking studies.

Multiobjective evolutionary algorithms (MOEAs) have been widely used in community detection in recent years. However, most of the existing MOEA-based ones adopted the same search strategies for all nodes and ignored the differences between the nodes. In fact, the nodes in a complex network have different structural characteristics and are of different importance during the search process of the community detection problem. To this end, in this article, a node classification-based search scheme is first proposed, where different kinds of nodes are searched in different ways. To be specific, the nodes in the network are classified into two types of nodes, candidate central (CC) nodes and noncentral (NC) nodes, by mapping the nodes into a structural similarity-based embedding space. The CC nodes are likely to be the centers of communities, and the rough structure can be searched quickly through activating the CC nodes. Then, the NC nodes are assigned to the communities with the activated central nodes. Based on the proposed scheme, a node classification-based MOEA named NCMOEA is then proposed. In NCMOEA, a mixed representation is designed to effectively encode the two different kinds of nodes. In addition, corresponding genetic operators are then suggested to search the two categories of nodes in different ways. Furthermore, an initialization strategy is also designed for initializing the population with high quality and good diversity. The experimental results on 15 real-world networks and several synthetic networks demonstrate the superiority of the proposed NCMOEA over nine representative algorithms for community detection.

Dynamic community detection, which is capable of revealing changes in community structure over time, has garnered increasing attention in research. While evolutionary clustering methods have proven to be effective in tackling this issue, they often have a tendency to favor what are referred to as elite solutions, inadvertently neglecting the potential value of non-elite alternatives. Although elite solutions can ensure population convergence, they may result in negative population migration due to the lack of diversity when the network changes. In contrast, when the network undergoes changes, non-elite solutions could better adapt to the changed network, thereby can help the algorithm find accurate community structures in the new environment. To this end, we propose a diversified population migration strategy that consists of two-stages, i.e., solution selection and solution migration. In the first stage, we use elite solutions not only to ensure convergence but also non-elite solutions to maintain diversity and cope with network changes. In the second stage, the migration solutions are refined by using incremental changes between the two consecutive snapshots of networks. Based on the proposed strategy, we suggest a diversified population migration-based multiobjective evolutionary algorithm named DPMOEA. In DPMOEA, we design new genetic operators that utilize incremental changes between networks to make the population evolve in the right direction. Our experimental results demonstrate that the proposed method outperforms state-of-the-art baseline algorithms and can effectively solve the dynamic community detection problem.

Local community detection (LCD) aims to find a community containing a given starting node using local information, which becomes a hot research topic in the area of community detection. Most mainstream studies in this field only consider network topology information, neglecting node attribute information. Some studies consider both, but these methods are limited by inflexible predefined structures. When the local community topology structure is unclear or the node attributes are heterogeneous, the performance of these methods decreases. In this paper, in order to better balance topological and attribute information, we propose an evolutionary multitasking local community detection (EMLCD) framework to solve the problem of local community detection on attributed networks. In EMLCD, the LCD problem is formulated as a multitasking optimization problem with two tasks, that is, Top-Task detects local community with tight topological connections, and Att-Task detects local community with homogeneous node attributes. The advantage of using evolutionary multitasking to detect local community on attributed networks is that the correlation between these tasks can be used to enhance detection performance. A novel knowledge transfer strategy is proposed to facilitate positive transfer of knowledge between the two tasks by leveraging the complementarity of the topology information and attribute information. Finally, a solution selection strategy is proposed to guide the decision maker in selecting the ideal solution from the set of solutions generated by the two tasks. Experimental results show that the proposed framework is very effective in solving the problem of local community detection on attributed networks.

Understanding the protein structures is invaluable in various biomedical applications, such as vaccine development. Protein structure model building from experimental electron density maps is a time-consuming and labor-intensive task. To address the challenge, machine learning approaches have been proposed to automate this process. Currently, the majority of the experimental maps in the database lack atomic resolution features, making it challenging for machine learning-based methods to precisely determine protein structures from cryogenic electron microscopy density maps. On the other hand, protein structure prediction methods, such as AlphaFold2, leverage evolutionary information from protein sequences and have recently achieved groundbreaking accuracy. However, these methods often require manual refinement, which is labor intensive and time consuming. In this study, we present DeepTracer-Refine, an automated method that refines AlphaFold predicted structures by aligning them to DeepTracers modeled structure. Our method was evaluated on 39 multi-domain proteins and we improved the average residue coverage from 78.2 to 90.0% and average local Distance Difference Test score from 0.67 to 0.71. We also compared DeepTracer-Refine with Phenixs AlphaFold refinement and demonstrated that our method not only performs better when the initial AlphaFold model is less precise but also surpasses Phenix in run-time performance.

The EMDataResource Ligand Model Challenge aimed to assess the reliability and reproducibility of modeling ligands bound to protein and protein/nucleic-acid complexes in cryogenic electron microscopy (cryo-EM) maps determined at near-atomic (1.9-2.5 Å) resolution. Three published maps were selected as targets: E. coli beta-galactosidase with inhibitor, SARS-CoV-2 RNA-dependent RNA polymerase with covalently bound nucleotide analog, and SARS-CoV-2 ion channel ORF3a with bound lipid. Sixty-one models were submitted from 17 independent research groups, each with supporting workflow details. We found that (1) the quality of submitted ligand models and surrounding atoms varied, as judged by visual inspection and quantification of local map quality, model-to-map fit, geometry, energetics, and contact scores, and (2) a composite rather than a single score was needed to assess macromolecule+ligand model quality. These observations lead us to recommend best practices for assessing cryo-EM structures of liganded macromolecules reported at near-atomic resolution.

Local overlapping community detection (LOCD), which aims to discover the multiple communities containing a given starting node, becomes a hot research problem in community detection area. However, existing LOCD methods only consider the quality of each local community containing the starting node, while ignoring the information between the multiple local communities. To this end, in this paper, we first formulate a novel multi-objective model to better assess the quality of local overlapping communities, where two novel objectives are suggested by fully utilizing the information within and between the multiple local communities. Afterward, we propose a tri-division representation-based multi-objective evolutionary algorithm (TDR-MOEA) for effectively finding multiple local overlapping communities for a given node under the formulated multi-objective model. In TDR-MOEA, a tri-division individual representation scheme is designed to effectively encode and decode the local overlapping communities. Based on the representation, two novel crossover and mutation operators are then suggested. In addition, an effective population initialization strategy is designed to obtain an initial population with better diversity and convergence. The effectiveness of the proposed multi-objective model and TDR-MOEA is verified on several synthetic networks and real-world networks. The experimental results show that TDR-MOEA is superior to state-of-the-art algorithms for solving LOCD.

Although some pyroptosis-related (PR) prognostic models for cancers have been reported, pyroptosis-based features have not been fully discovered at the single-cell level in hepatocellular carcinoma (HCC). In this study, by deeply integrating single-cell and bulk transcriptome data, we systematically investigated significance of the shared pyroptotic signature at both single-cell and bulk levels in HCC prognosis. Based on the pyroptotic signature, a robust PR risk system was constructed to quantify the prognostic risk of individual patient. To further verify capacity of the pyroptotic signature on predicting patients' prognosis, an attention mechanism-based deep neural network classification model was constructed. The mechanisms of prognostic difference in the patients with distinct PR risk were dissected on tumor stemness, cancer pathways, transcriptional regulation, immune infiltration and cell communications. A nomogram model combining PR risk with clinicopathologic data was constructed to evaluate the prognosis of individual patients in clinic. The PR risk could also evaluate therapeutic response to neoadjuvant therapies in HCC patients. In conclusion, the constructed PR risk system enables a comprehensive assessment of tumor microenvironment characteristics, accurate prognosis prediction and rational therapeutic options in HCC.

Understanding the structures of proteins has numerous applications, such as vaccine development. It is a slow and labor-intensive task to manually build protein structures from experimental electron density maps, therefore, machine learning approaches have been proposed to automate this process. However, most of the experimental maps are not atomic resolution, so they are insufficient for computer vision-based machine learning methods to precisely determine the protein structure. On the other hand, methods that utilize evolutionary information from protein sequences to predict structures, like AlphaFold2, have recently achieved groundbreaking accuracy but often require manual effort to refine the results. We propose DeepTracer-Refine, an automated method to refine AlphaFold structures by aligning them to DeepTracer’s predicted structure. We tested our method on 39 multi-domain proteins and we improved the average residue coverage from 78.2% to 90.0% and average lDDT score from 0.67 to 0.71. We also compared DeepTracer-Refine against another method, Phenix’s AlphaFold refinement, to demonstrate that our method not only performs better when the initial AlphaFold model is less precise but also exceeds Phenix in run-time performance.

MOTIVATION:In recent years, the end-to-end deep learning method for single-chain protein structure prediction has achieved high accuracy. For example, the state-of-the-art method AlphaFold, developed by Google, has largely increased the accuracy of protein structure predictions to near experimental accuracy in some of the cases. At the same time, there are few methods that can evaluate the quality of protein complexes at the residue level. In particular, evaluating the quality of residues at the interface of protein complexes can lead to a wide range of applications, such as protein function analysis and drug design. In this paper, we introduce a new deep graph neural network-based method ComplexQA, to evaluate the local quality of interfaces for protein complexes by utilizing the residue-level structural information in 3D space and the sequence-level constraints.RESULTS:We benchmark our method to other state-of-the-art quality assessment approaches on the HAF2 and DBM55-AF2 datasets (high-quality structural models predicted by AlphaFold-Multimer), and the BM5 docking dataset. The experimental results show that our proposed method achieves better or similar performance compared with other state-of-the-art methods, especially on difficult targets which only contain a few acceptable models. Our method is able to suggest a score for each interfac e residue, which demonstrates a powerful assessment tool for the ever-increasing number of protein complexes.AVAILABILITY:https://github.com/Cao-Labs/ComplexQA.git. Contact: caora@plu.edu.

Cryo-electron microscopy (cryo-EM) allows a macromolecular structure such as protein-DNA/RNA complexes to be reconstructed in a three-dimensional coulomb potential map. The structural information of these macromolecular complexes forms the foundation for understanding the molecular mechanism including many human diseases. However, the model building of large macromolecular complexes is often difficult and time-consuming. We recently developed DeepTracer-2.0, an artificial-intelligence-based pipeline that can build amino acid and nucleic acid backbones from a single cryo-EM map, and even predict the best-fitting residues according to the density of side chains. The experiments showed improved accuracy and efficiency when benchmarking the performance on independent experimental maps of protein-DNA/RNA complexes and demonstrated the promising future of macromolecular modeling from cryo-EM maps. Our method and pipeline could benefit researchers worldwide who work in molecular biomedicine and drug discovery, and substantially increase the throughput of the cryo-EM model building. The pipeline has been integrated into the web portal https://deeptracer.uw.edu/.