MOTIVATION: The increasing prevalence of antibiotic-resistant bacteria has intensified the demand for novel antimicrobial agents. Antimicrobial peptides (AMPs) have emerged as promising alternatives, yet their identification or classification remains challenging due to the lack of multi-perspective information, insufficient feature representation learning, and monocular data modalities. RESULTS: In this paper, we propose a dual diffusion model-based representation learning framework for classifying AMPs, which effectively integrates both peptide sequence and structure information to address existing issues for the task. Specifically, our approach utilizes a multi-view feature construction module, which encodes peptide sequences and structures from distinctive perspectives, deriving initial feature representations with enriched biological semantics. To enhance representation learning, the proposed framework leverages both diffusion models for sequence and structure information respectively to effectively capture complex semantics from dual modalities. In addition, both single-modal and dual-modal contrastive learning are used to further advance the representation learning. Results of comprehensive experiments demonstrate that our model outperforms existing methods for the task of AMPs classification, providing a feasible solution to accelerating the discovery of novel antimicrobial agents. AVAILABILITY OF IMPLEMENTATION: The data and source codes are available in GitHub at https://github.com/kww567upup/DDM.

Drug repositioning is an efficient drug discovery method for identifying associations between present drugs and new diseases, offering considerable development time and cost savings. Although existing methods have been widely applied, they fail to fully capture the complex semantics between drugs and diseases, and are deficient in terms of model interpretability. In this paper, we propose a novel method using a hierarchical attention mechanism aggregating meta-path information for drug-disease association prediction (MPHAM), aiming at effectively integrating heterogeneous information from various sources to enhance prediction accuracy and model interpretability. First, considering the wide range of biological interactions between drugs and diseases, we construct a heterogeneous information network (HIN) to utilize data on drugs, proteins, and diseases. Then, we introduce a meta-path-based feature fusion strategy designed to effectively capture the complex semantics between nodes in the network. By defining meta-paths of multiple lengths and types, information about different relationship types is systematically integrated to generate high-quality node feature representations. Furthermore, the feature fusion strategy incorporates a multi-layer attention mechanism that dynamically assigns weights to the contributions of various meta-paths in the feature aggregation process, significantly improving the model's capacity to capture important semantic information. Experimental results demonstrate that MPHAM can effectively predict drug-disease association by integrating complex meta-path information, and the prediction accuracy is better than five state-of-the-art methods. The case studies of three classical drugs further demonstrate the more accurate predictive performance of MPHAM in drug-candidate disease association prediction.

Transcription functions as a pivotal biological process in cell biology, which is required to complete the binding of transcription factors (TFs) to transcription factor binding sites (TFBSs) on the DNA. Accurate prediction of TFBSs can provide great potential to regulate the expression of interested genes, which can facilitate exploration of new drugs and treatment for diseases. Although many deep learning-based models have been proposed for predicting TFBSs, existing models still have problems, including the use of convolutional processing of DNA sequences that loses information about the DNA double helix structure and fails to adequately account for the stereoscopic structure of DNA shape data in three dimensions. In this paper, we propose a novel model called DeepCTMS, in which both sequence features and shape features of DNA slices are effectively fused to derive high-quality representations for the task of TFBS prediction. A sequence feature processing module is first used to extract the DNA double helix structure features of DNA slices. The three-dimensional features of DNA shape data are extracted by employing a convolutional triple attention (CTA) module for the shape data of a DNA slice. Finally, a multi-scale fusion feature processing (MSFFP) module is used to fuse sequence features and shape features to obtain representations with significantly aligned semantics of both features. Ablation experiments, t-SNE visual analysis, and cross-cell line validation results demonstrate that DeepCTMS consistently outperforms benchmark models on prediction performance and generalization ability on 165 ChIP-seq datasets.

Machine Learning (ML) demonstrates dermatologist level accuracy in skin cancer diagnosis, yet its practical adoption is constrained by data silos and privacy issues. While Federated Learning (FL) addresses these limitations, it remains susceptible to data heterogeneity and gradient leakage attacks. To overcome these challenges, we introduce a privacy-preserving FL framework tailored for encrypted dermoscopic image analysis. Our proposed framework integrates a Fully Homomorphic Encryption (FHE)-enabled variant of Stochastic Controlled Averaging (SCA), enhancing model convergence with Non-IID data. To further minimize computational and communication overhead, we develop a layer-wise Packed FHE (PFHE) approach that improves the efficiency of encrypted model aggregation. Moreover, we design a lightweight, FHE-Friendly Deep Neural Network (DNN) optimized for encrypted inference. This architecture incorporates a DO-EncConv module specifically engineered to balance inference efficiency and precision within FHE computational constraints. Experimental results on the HAM10000 and ISIC2019 datasets confirm the effectiveness of our proposed framework, demonstrating F1-Score improvements of 2.2% and 4.0%, respectively, over baseline FL approaches. Additionally, our method achieves communication overhead reductions of 94.85% and 93.48%, while encrypted inference is performed in approximately 17.8 seconds per sample, with less than 2% accuracy degradation compared to centralized plaintext models. These outcomes underscore the framework's practicality and effectiveness for secure, scalable clinical deployment.

Influenza-like illness (ILI) forecasting is crucial for effective public health intervention, but existing models often fail to capture the complex temporal and semantic patterns inherent in epidemic data. Traditional statistical techniques and even advanced deep learning methods predominantly leverage numerical time series data, thereby overlooking contextual medical and epidemiological insights that could enhance the performance. Recent progress in large language models (LLMs) has illustrated their exceptional effectiveness in integrating semantic understanding into natural language processing tasks within the medical context. Motivated by these developments, we propose SMILE (Semantic Multi-scale Integration and LLM-Enhanced network), a novel multi-modal forecasting framework designed to integrate LLM-derived semantic features with multi-scale temporal analysis. Built upon TimeMixer architecture, SMILE introduces an automatic semantic feature extraction system using LLMs, adaptive fusion mechanisms for integrating textual and temporal data, and demonstrates robust performance improvements. Extensive experiments on ILI and benchmark datasets confirm that SMILE significantly outperforms state-of-the-art forecasting methods, highlighting the value of incorporating semantic context into time series disease prediction.

Identifying drug–target interactions (DTIs) is critical for discovery and repurposing. Existing deep models often ignore frequency-domain structure and rely on global pooling that dilutes binding-critical signals. We present TriPeakDTI, which couples adaptive frequency-domain analysis with a peak-guided attention mechanism. Using the discrete cosine transform (DCT), TriPeakDTI learns per-modality spectral weights for drugs and targets to capture multi-scale patterns—low frequencies for global conformations and high frequencies for local flexibility—during single-modal extraction. A bidirectional, peak-guided fusion module preserves token-level interaction evidence via peak detection, cross-modal alignment, and strength-aware aggregation, preventing information washout. Across three benchmark datasets, TriPeakDTI outperforms seven state-of-theart methods.

Accurate prediction of phage-host interactions remains a fundamental challenge that impedes the clinical deployment of phage therapy. Current graph neural network-based methods rely on superficial sequence features, failing to adequately integrate deep biological semantic information. In this work, we present DSHMGAT (Dual-Stream Hierarchical Mixed-Routing Graph Attention Network), a novel framework that incorporates agent-generated semantic embeddings to mitigate this semantic deficiency. By employing a ReAct-driven agent with function calling capabilities to access domain-specific databases, our approach reduces hallucinations inherent in biological LLM applications. DSHMGAT adopts a dualstream hierarchical graph neural network that simultaneously captures genomic sequences and biological semantic representations, enabling effective cross-modal information integration. Inspired by mixture-of-experts architectures, we develop a mixed-routing attention mechanism that improves learning flexibility through dynamic weight allocation between routing heads and shared heads. DSHMGAT achieves an AUC of 0.9486 on the benchmark dataset which outperforms established approaches in our comparative analysis. Ablation experiments reveal that the observed improvements can be attributed to the combined effects of multimodal fusion, cross-modal interaction and mixed-routing mechanism.

The task of nanobody paratope prediction aims to identify the residues on a nanobody that specifically bind to an antigen, which is crucial for understanding the mechanisms of an effective nanobody agent, providing significant implications for drug development accordingly. Although various methods have been proposed based on different strategies, existing methods still face several challenges: overlooking nanobody-specific structural characteristics, neglecting optimization of high-dimensional features extracted from antibody language models, and disregarding the antigen information crucial for nanobody-specific binding. To address these challenges, we propose NanoFADIL, a novel nanobody paratope prediction model that employs a fused attention mechanism combining channel attention and one-dimensional spatial attention to optimize feature representations adaptively, and uses residue-level distance supervision to guide the learning of nanobody-antigen interaction patterns. Specifically, the existing nanobody paratope dataset is augmented by incorporating the sequences of corresponding antigens, allowing the prediction of antigen-specific paratopes on nanobodies. Nanobodies are encoded by the pretrained antibody language model IgT5 to derive meaningful feature representations. High-dimensional nanobody representations are adaptively reweighted across feature channels and residue positions by our fused attention mechanism that combines channel-wise and one-dimensional spatial attention, thereby enabling the model to focus on informative features relevant to nanobody paratope prediction and contextually important residues. During training, residue-level distances between nanobodies and antigens are transformed into supervision signals, guiding the model to learn meaning-ful nanobody-antigen interaction patterns and enhancing the model's predictive capability. Experimental results demonstrate that NanoFADIL achieves superior performance compared to existing methods.

The critical role of nutrients in intervening in human diseases has made a deeper exploration of microbial metabolic processes greatly needed. Bacteriophages serve as key regulators in maintaining microecological balance. Existing studies have revealed that nutrients promote bacteriophage production in gut bacteria, subsequently affecting bacterial abundance and influencing disease development. However, current research lacks systematic integration of the relationships among bacteriophages, nutrition, gut bacteria, and diseases, which partially limits our understanding of the intervention mechanisms between nutrition and diseases. To address this, we constructed a novel knowledge graph, KGNBVD, focusing on the interconnections among dietary nutrition, gut bacteria, bacteriophages, and human diseases by harnessing large-scale biomedical literature and the promptdriven capabilities of large language models (lLMs). Additionally, to elucidate the mechanistic role of bacteriophages in linking nutrients and human diseases, we proposed a multi-view fusion approach that fully explore the features of KGNBVD. Extensive experimental results demonstrated that our method outperforms state-of-the-art methods in predicting four-entity interactions with AUC values of 96.20 % and$\mathbf{F 1}$scores of$\mathbf{8 9. 7 2 \%}$. Moreover, results of case study validated that our method can identify high-probability candidate entities for previously unseen entity interactions, effectively predicting potential interactions between nutrition and human diseases.

Salmonella infections pose a significant global public health threat, and the emergence of multi-drug resistant strains has left traditional antibiotic agents in a predicament of ineffectiveness. Since the development of traditional antibiotic drugs is costly and time-consuming, the global research and development of novel antibiotics has nearly stagnated at present. Although several generative molecular design models have been proposed, their application in generating antibiotics targeting specific pathogens suffers from undesirable antibacterial efficacy and poor diversity due to limited training data. In this paper, we propose an effective Variational AutoEncoder (VAE) framework for generating antibiotic molecules targeting Salmonella bacteria, in which a triple-stage training strategy is utilized to address the issues of poor efficacy and low diversity for existing molecule generation models. More specifically, a set of all available small molecule drugs is first used to pre-train the VAE-based framework to capture the general semantics contained in various types of small molecules. Then, a mediumsize set of all antibiotic molecules is employed to fine-tune the pre-trained VAE framework to learn the fine-grained semantics among antibiotic agents. Finally, a small-size set of all antibiotics treating Salmonella pathogens is utilized to optimize further the generated molecules with specific anti-Salmonella activity. In addition, introducing several key molecular properties as conditional input can effectively guide the VAE model to generate molecules with ideal absorption and metabolic properties. Results of comprehensive experiments show that the proposed framework is able to generate highly effective and structurally diverse antibacterial candidate molecules, providing a feasible solution to the development of antibiotics against drug-resistant Salmonella pathogens.

The prediction of protein-peptide binding residues is to identify the protein residues that can bind to peptides, which is essential for uncovering protein functions and supporting peptide-based drug development. Despite the progress made by previous methods, several challenges remain, including the neglect of protein domain information, the underutilization of protein language models, and the issue of data imbalance. In this paper, we propose PepGFD, a novel model based on a gated fusion mechanism and domain-guided feature optimization to predict protein-peptide binding residues. Specifically, PepGFD employs a gated fusion mechanism to fuse complementary features extracted from two protein language models, ESM-2 and ProtT5, to construct protein node features. Meanwhile, multiple properties are extracted from the protein structure to construct protein edge features. To guide the model in identifying key functional residues on the protein, we incorporate protein domain information to optimize protein node features. Subsequently, we employ multiple attention mechanisms to process protein features and peptide features separately, while effectively capturing the semantics among interactions between them. Additionally, a dualstrategy contrastive learning (DSCL) approach is used to address the issue of data imbalance, thereby enabling the model to better distinguish between binding and non-binding residues. Experimental results on two widely used benchmark datasets demonstrate that PepGFD achieves superior performance compared to existing models.

Identifying phage-host interactions (PHIs) is a crucial step in developing phage therapy, which is the promising solution to addressing the issue of antibiotic resistance in superbugs. However, the lifestyle of phages, which strongly depends on their host for life activities, limits their cultivability, making the study of predicting PHIs time-consuming and labor-intensive for traditional wet lab experiments. Although many deep learning (DL) approaches have been applied to PHIs prediction, most DL methods are predominantly based on sequence information, failing to comprehensively model the intricate relationships within PHIs. Moreover, most existing approaches are limited for sub-optimal performance, due to the potential risk of overfitting induced by the highly data sparsity in the task of PHIs prediction. In this study, we propose a novel approach called MI-RGC, which introduces mutual information for feature augmentation and employs regional graph convolution to learn meaningful representations. Specifically, MI-RGC treats the presence status of phages in environmental samples as random variables, and derives the mutual information between these random variables as the dependency relationships among phages. Consequently, a mutual information-based heterogeneous network is construted as feature augmentation for sequence information of phages, which is utilized for building a sequence information-based heterogeneous network. By considering the different contributions of neighboring nodes at varying distances, a regional graph convolutional model is designed, in which the neighboring nodes are segmented into different regions and a regional-level attention mechanism is employed to derive node embeddings. Finally, the embeddings learned from these two networks are aggregated through an attention mechanism, on which the prediction of PHIs is condcuted accordingly. Experimental results on three benchmark datasets demonstrate that MI-RGC derives superior performance over other methods on the task of PHIs prediction.

Bacterial resistance has emerged as one of the greatest threats to human health, and phages have shown tremendous potential in addressing the issue of drug-resistant bacteria by lysing host. The identification of phage-host interactions (PHI) is crucial for addressing bacterial infections. Some existing computational methods for predicting PHI are suboptimal in terms of prediction efficiency due to the limited types of available information. Despite the emergence of some supporting information, the generalizability of models using this information is limited by the small scale of the databases. Additionally, most existing models overlook the sparsity of association data, which severely impacts their predictive performance as well. In this study, we propose a dual-view sparse network model (DSPHI) to predict PHI, which leverages logical probability theory and network sparsification. Specifically, we first constructed similarity networks using the sequences of phages and hosts respectively, and then sparsified these networks, enabling the model to focus more on key information during the learning process, thereby improving prediction efficiency. Next, we utilize logical probability theory to compute high-order logical information between phages (hosts), which is known as mutual information. Subsequently, we connect this information in node form to the sparse phage (host) similarity network, resulting in a phage (host) heterogeneous network that better integrates the two information views, thereby reducing the complexity of model computation and enhancing information aggregation capabilities. The hidden features of phages and hosts are explored through graph learning algorithms. Experimental results demonstrate that mutual information is effective information in predicting PHI, and the sparsification procedure of similarity networks significantly improves the model's predictive performance.

MOTIVATION: The widespread use of antibiotics has led to the emergence of resistant pathogens. Antimicrobial peptides (AMPs) combat bacterial infections by disrupting the integrity of cell membranes, making it challenging for bacteria to develop resistance. Consequently, AMPs offer a promising solution to addressing antibiotic resistance. However, the limited availability of natural AMPs cannot meet the growing demand. While deep learning technologies have advanced AMP generation, conventional models often lack stability and may introduce unforeseen side effects. RESULTS: This study presents a novel denoising VAE-based model guided by desirable physicochemical properties for AMP generation. The model integrates key features (e.g. molecular weight, isoelectric point, hydrophobicity, etc.), and employs position encoding along with a Transformer architecture to enhance generation accuracy. A customized loss function, combining reconstruction loss, KL divergence, and property preserving loss ensure effective model training. Additionally, the model incorporates a denoising mechanism, enabling it to learn from perturbed inputs, thus maintaining performance under limited training data. Experimental results demonstrate that the proposed model can generate AMPs with desirable functional properties, offering a viable approach for AMP design and analysis, which ultimately contributes to the fight against antibiotic resistance. AVAILABILITY AND IMPLEMENTATION: The data and source codes are available both in GitHub (https://github.com/David-WZhao/PPGC-DVAE) and Zenodo (DOI 10.5281/zenodo.14730711).

Antimicrobial peptides (AMPs) are crucial in addressing the global crisis of bacterial resistance. However, there are still significant limitations in existing methods on de novo AMPs design, especially in designing AMPs with desirable physicochemical properties for specific bacterial pathogens. In this study, we propose a novel generative framework for designing pathogen-targeted antimicrobial peptides with programmable physicochemical properties. More specifically, a conditional Variational Autoencoder is first pretrained for generating AMPs with editable physicochemical properties. We then develop a conditional diffusion model to learn hidden representations of AMPs for targeting pathogens of interest, and construct corresponding MIC predictors for specific bacterial strains. Through comprehensive simulation experiments, we demonstrate that the proposed framework outperforms most existing models in terms of antimicrobial efficacy against specific bacterial targets. Moreover, through systematic screening and analysis, we have identified two star AMPs for each of the two target bacterial species (i.e., E. coli or S. aureus), both of which exhibit excellent performance in antibacterial activity, hemolytic properties, toxicity profiles, etc. Overall, this study provides the key technological support for developing next-generation intelligent platforms for antimicrobial agents design.

It is commonly known that food nutrition is closely related to human health. The complex interactions between food nutrients and diseases, influenced by gut microbial metabolism, present challenges in systematizing and practically applying knowledge. To address this, we propose a method for extracting triples from a vast amount of literature, which is used to construct a comprehensive knowledge graph on nutrition and human health. Concurrently, we develop a query-based question answering system over our knowledge graph, proficiently addressing three types of questions. The results show that our proposed model outperforms other state-of-art methods, achieving a precision of 0.92, a recall of 0.81, and an F1 score of 0.86 in the nutrition and disease relation extraction task. Meanwhile, our question answering system achieves an accuracy of 0.68 and an F1 score of 0.61 on our benchmark dataset, showcasing competitiveness in practical scenarios. Furthermore, we design five independent experiments to assess the quality of the data structure in the knowledge graph, ensuring results characterized by high accuracy and interpretability. In conclusion, the construction of our knowledge graph shows significant promise in facilitating diet recommendations, enhancing patient care applications, and informing decision-making in clinical research.

Due to the abuse of antibiotics, some pathogenic bacteria have developed resistance to most antibiotics, leading to the emergence of antibiotic-resistant superbugs. Therefore, researchers resort to phage therapy for bacterial infections. For phage therapy, the fundamental step is to accurately identify phage-host interactions. Although various methods have been proposed, the existing methods suffer from the following two shortcomings: 1) they fail to make full use of genetic information including both genome and protein sequence of phages; 2) host specificity of phages is not explicitly utilized when learning representations of phages and bacteria. In this paper, we present an efficient computational method called PHISGAE for predicting phage-host interactions, in which the host specificity is explicitly employed. Firstly, initial phage-phage connections are efficiently constructed via utilizing phage genome and protein sequence. Then, the refined heterogeneous network is derived by applying K-nearest neighbor strategy, keeping relatively more meaningful local semantics among phages and bacteria. Finally, a host specificity-aware graph autoencoder is proposed to learn high-quality representations of phages and bacteria for predicting phage-host interactions. Experimental results show that PHISGAE outperforms the state-of-the-art methods on predicting phage-host interactions at both species level and genus level (AUC values of 94.73% and 96.32%, respectively). Moreover, results of case study demonstrate that PHISGAE is able to identify candidate hosts with high probability for previously unseen phages identified from metagenomics, effectively predicting potential phage-host interactions in real-world applications.

Microbial-disease association prediction has always been a frontier research direction in bioinformatics, which includes two tasks: predicting association relationships and association categories (Decrease, Increase). At present, The study methods based on statistical analysis rely heavily on biological prior knowledge, the traditional microbial experiments are time-consuming and costly. In this paper, we propose a novel model that predict microbe-disease association based on an enhanced relational graph convolutional network. Firstly, we construct a heterogeneous network containing microbial abundance changes and disease associations, microbial similarity, and disease similarity. Then, this paper proposes a feature difference enhancement module based on graph similarity, which aims to enhance the difference of feature representation between different association categories. It is fused with the relational graph convolutional network to form an enhanced relational graph convolutional network and complete feature coding work. Thus, the multi-category feature representation between nodes can be effectively extracted and the influence of edges on the graph structure can be considered. Finally, a neural network was used to complement nonlinear feature representation. The experimental results show that the proposed model can efficiently predict the association categories between microbial abundance changes and diseases.

The biomedical literature suggests that drug interventions can affect microbial communities of the human body and thereby treat disease. Screening of microbe-drug associations is therefore a key research topic in the field of antimicrobial drug development and drug repositioning. Existing models have not adequately integrated the multimodal representations of biomedical entities and the microbe-drug association networks from different sources. In addition, the positive negative sample imbalance of the heterogeneous networks and the sparsity of the association matrices decrease the accuracy of the prediction results. In this paper, we proposed a multimodal joint representation framework (MJRFMDA) for microbe-drug association prediction. First, we represented microbial and drug features through multimodal attributes. Secondly, we aggregated neighbor information and updated node attributes through GCN and GAT in microbe-drug association networks. Then we aggregated the embedded representations of multimodal networks through a joint representation layer based on a graph-level attention mechanism. Finally, we employed neural inductive matrix completion to capture the nonlinear associations between microbes and drugs, and reconstructed the association score matrix. The experimental results show that MJRFMDA outperforms the six selected state-of-the-art models.

Antimicrobial peptides (AMPs) hold significant promise in antibacterial and anticancer research, offering a crucial solution to the escalating issue of antibiotic resistance and paving the way for developing novel strategy for cancer treatments. Previously, experimental methods were used to identify AMPs. However, there is now a shift towards developing computational approaches to predict AMPs accurately, greatly reducing the time and effort required for experimental identification. Past computation methods for predicting AMPs have either focused on feature encoding and extraction or emphasized neural network design. Unfortunately, relying on a single method or feature can lead to the oversight of important information. In this study, we present AMPpred-DLFF, a novel computational method for identifying AMPs. This method initially employs ESM-2 to generate residue embedding representations, captures the spatial relationships between residues, and constructs a graph to integrate peptide data. Subsequently, a graph attention network extracts features from the graph data. Concurrently, multiple encoding techniques are used to extract residue features from sequence data, which are then processed by a convolutional neural network. Finally, the feature representations from both modules are combined, and the peptide sequence is predicted as an antimicrobial peptide through a fully connected layer. Performance assessments on several benchmark datasets demonstrate that AMPpred-DLFF outperforms current leading methods, underscoring the effectiveness of multi-view fusion in enhancing AMP prediction accuracy.

Accurate prediction of phage-host interactions (PHIs) is crucial for phage therapy and antimicrobial substitution research. Although many methods have been proposed for the task of PHIs prediction, most of them fail to employ the biological knowledge in phage-host interaction, such as the receptor-binding protein (RBP) of phages plays an essential role in the binding mechanism between phages and hosts, leading to undesirable prediction performance. In this study, we propose a novel PHIs prediction model in which the RBP data is explicitly employed. Specifically, RBP information and whole protein sequence for each phage information are integrated by the cross-attention mechanism, by which global and local semantics contained in phages are effectively captured. For hosts, the protein sequences are used to extract the features accordingly. Then, the final representation is obtained by combining the phage features and host features, on which a MLP is utilized to derive the predicted probability of interaction between the input phage-host pair. Comprehensive experiments are conducted and the results demonstrate the effectiveness of the proposed method.

With the widespread use of antibiotics, many pathogens have gradually developed resistance. However, antimicrobial peptides (AMPs) exert their effects by directly disrupting the integrity of the cell membrane, making it difficult for bacteria to develop resistance through genetic mutations. This makes AMPs an important agent in addressing the issue of antibiotics resistance globally. Since the naturally obtained AMPs are limited in quantity and insufficient to meet diverse demands, the emergence of deep learning has led to breakthroughs in the field of synthetic AMPs. However, models that generate AMPs unconditionally often suffer from poor interpretability and potential side effects. In this paper, we propose a novel model for AMPs generation via physico-chemical property-guided VAE. This model comprehensively considers various physicochemical properties such as molecular weight, isoelectric point, hydrophobicity, and etc. By integrating positional encoding and a transformer structure, it effectively learns the generative process of AMPs with required physicochemical properties. Moreover, an innovative loss function is designed, which combines the reconstruction loss, KL divergence, and attribute mapping loss to effectively train the conditional VAE. Experimental results demonstrate that this model is able to generate AMPs with desirable functional characteristics, providing a feasible solution to the design and analysis of AMPs.

The increasing prevalence of antibiotic-resistant bacteria has made novel antimicrobial agents greatly needed. Antimicrobial peptides (AMPs) have emerged as promising alternatives, but their identification and classification remain challenging. In this paper, we propose a dual-modal contrastive learning framework for the task of AMPs classification, in which both sequence and structural information of peptides are fully used and effectively integrated. Specifically, a multi-view data processing module aims to encode peptides from multiple perspectives, deriving initial sequence-based features and structure-based features. For better feature learning, a Transformer architecture is conducted on sequence-based features, while a Graph Attention Network (GAT) module is conducted on structure-based features. Moreover, we introduce a dual-modal contrastive learning strategy to enhance the model’s ability for discriminating AMPs and non-AMPs, and to align semantics contained in different modalities. We conduct comprehensive experiments and the results demonstrate the effectiveness of our method on the task of AMPs classification.

Accurately predicting protein-ligand affinity is one of the critical steps in the field of drug design. While deep learning approaches have shown great potential, existing methods based on sequence and 3D structural information still face challenges in capturing the spatial structure information and molecular bonding interactions between proteins and ligands. The advantage of molecular mechanics lies in its ability to account for complex molecular interactions, including electrostatic interactions and van der Waals forces. These interactions are key factors in determining the binding affinity between the pair of ligand and protein. Therefore, this study proposes a molecular mechanics-based heterogeneous graph attention neural network for predicting protein-ligand binding affinity. More specifically, various molecular mechanics features and atomic type features are first collected. Then, the heterogeneous graph is constructed for each pair of protein and ligand, in which the nodes are distinguished by atomic types, and the edges are categorized into covalent and non-covalent bonds. For graph representation learning, the molecular mechanics-guided aggregation mechanism is introduced to learn the meaningful constraints contained in the protein-ligand complexes. Finally, the representations of protein-ligand complexes are derived, on which the protein-ligand affinity is predicted accordingly. Experimental results show that the proposed model outperforms selected baselines on the task of protein-ligand affinity prediction.

Concept-based approaches have emerged as a promising solution to enhance search accuracy in biomedical information retrieval (BIR). These approaches aim to bridge the semantic gap between user queries and document content by mapping both to a shared conceptual space. However, existing methods often struggle with limited concept coverage and inaccurate concept mapping. Recent advancements in large language models (LLMs) have demonstrated their extensive concept coverage and ability to accurately connect text with relevant concepts, showing potential in addressing these challenges. Motivated by these findings, we propose Concept-based Biomedical Text Expansion (CBTE), a novel method that leverages LLMs to identify relevant concepts in both queries and documents. CBTE then uses these concepts to expand queries and documents. The relevance between expanded queries and documents is evaluated using the BM25 sparse retrieval framework. To validate CBTE’s effectiveness, we conducted comprehensive experiments on two well-known biomedical datasets, NFCorpus and Trec-Covid. The results indicate that CBTE significantly outperforms existing baselines.

MOTIVATION: The crisis of antibiotic resistance, which causes antibiotics used to treat bacterial infections to become less effective, has emerged as one of the foremost challenges to public health. Identifying the properties of antibiotic resistance genes (ARGs) is an essential way to mitigate this issue. Although numerous methods have been proposed for this task, most of these approaches concentrate solely on predicting antibiotic class, disregarding other important properties of ARGs. In addition, existing methods for simultaneously predicting multiple properties of ARGs fail to account for the causal relationships among these properties, limiting the predictive performance. RESULTS: In this study, we propose a causality-guided framework for annotating properties of ARGs, in which causal inference is utilized for representation learning. More specifically, the hidden biological patterns determining the properties of ARGs are described by a Gaussian Mixture Model, and procedure of causal representation learning is used to derive the hidden features. In addition, a causal graph among different properties is constructed to capture the causal relationships among properties of ARGs, which is integrated into the task of annotating properties of ARGs. The experimental results on a real-world dataset demonstrate the effectiveness of the proposed framework on the task of annotating properties of ARGs. AVAILABILITY AND IMPLEMENTATION: The data and source codes are available in GitHub at https://github.com/David-WZhao/CausalARG.

Computational drug repositioning can identify potential associations between drugs and diseases. This technology has been shown to be effective in accelerating drug development and reducing experimental costs. Although there has been plenty of research for this task, existing methods are deficient in utilizing complex relationships among biological entities, which may not be conducive to subsequent simulation of drug treatment processes. In this article, we propose a heterogeneous graph embedding method called HMLKGAT to infer novel potential drugs for diseases. More specifically, we first construct a heterogeneous information network by combining drug-disease, drug-protein and disease-protein biological networks. Then, a multi-layer graph attention model is utilized to capture the complex associations in the network to derive representations for drugs and diseases. Finally, to maintain the relationship of nodes in different feature spaces, we propose a multi-kernel learning method to transform and combine the representations. Experimental results demonstrate that HMLKGAT outperforms six state-of-the-art methods in drug-related disease prediction, and case studies of five classical drugs further demonstrate the effectiveness of HMLKGAT.

The crisis of antibiotic resistance has become a significant global threat to human health. Understanding properties of antibiotic resistance genes (ARGs) is the first step to mitigate this issue. Although many methods have been proposed for predicting properties of ARGs, most of these methods focus only on predicting antibiotic classes, while ignoring other properties of ARGs, such as resistance mechanisms and transferability. However, acquiring all of these properties of ARGs can help researchers gain a more comprehensive understanding of the essence of antibiotic resistance, which will facilitate the development of antibiotics. In this paper, the task of predicting properties of ARGs is modeled as a multi-task learning problem, and an effective subtask-aware representation learning-based framework is proposed accordingly. More specifically, property-specific expert networks and shared expert networks are utilized respectively to learn subtask-specific features for each subtask and shared features among different subtasks. In addition, a gating-controlled mechanism is employed to dynamically allocate weights to subtask-specific semantics and shared semantics obtained respectively from property-specific expert networks and shared expert networks, thus adjusting distinctive contributions of subtask-specific features and shared features to achieve optimal performance for each subtask simultaneously. Extensive experiments are conducted on publicly available data, and experimental results demonstrate the effectiveness of the proposed framework on the task of ARGs properties prediction.

The established links between nutrition and human health are widely acknowledged. Dietary nutrients play a crucial role in regulating gut microbial communities, influencing various human diseases. With a growing number of related studies, there’s a need to systematically organize these associations for coherent knowledge reasoning. However, due to the diverse and extensive nature of the knowledge landscape, significant challenges persist. To address this, we propose an approach using multimodal data and knowledge embeddings for effective knowledge reasoning in nutrition and human health. We create a comprehensive knowledge graph, KG4NH, covering dietary nutrition, gut microbiota, and human diseases. To ensure efficient knowledge representation, we employ knowledge embedding techniques to develop modality-specific encoders for structure, category, and description. Additionally, we introduce a mul-timodal fusion method to capture shared information across modalities. Our experimental results demonstrate the superiority of our approach over other state-of-the-art methods.

Antibiotic resistance event extraction involves the automated extraction of information related to antibiotic resistance mechanisms from a vast amount of biomedical literature. This can be achieved by utilizing natural language processing techniques. However, the distinctive characteristics of the biomedical field lead to various challenges for existing antibiotic resistance event extraction methods, such as limited labeling data, complex names of biomedical entities, and nesting and overlapping event structures. These factors make it challenging to apply the current processing methods for biomedical text to the task of antibiotic resistance event extraction. To address these challenges, we propose a cascade decoding approach for antibiotic resistance event extraction based on contrastive learning (CL-MA-CasEE). This approach achieves data augmentation by constructing two contrastive learning tasks, which combines entity type embedding and POS embedding to enrich the semantic information of word representations. Furthermore, it performs event type detection, event trigger extraction, and event argument extraction through using three cascade decoders to simulate the complex event structures. Based on experiments, we demonstrate that our method can effectively extract structured antibiotic resistance event information from biomedical literature, thereby improve the efficiency of event extraction tasks as well.

Detecting drug side effects is crucial in development of drugs. As publicly available biomedical data expands, researchers have devised numerous computational methods for predicting drug-side effect associations (DSAs). Among these, network-based approaches have gained significant attention in the biomedical field. However, the challenge of data scarcity poses a significant hurdle for existing DSAs prediction models. While various data augmentation methods have been created to solve the proble, most rely on random alterations to the original networks, neglecting the causality of DSAs’ existence, thus impacting the predictive performance negatively. In this paper, we introduce a counterfactual inference-based data augmentation method to enhance performance. First,a heterogeneous information network (HIN) is construct by integrating multiple biomedical data sources. We employ community detection on the HIN to preform a counterfactual inference-based method, deriving augmented links and an augmented HIN. Subsequently, we apply a meta-path-based graph neural network to obtain high-quality representations of drugs and side effects, enabling the prediction of DSAs. Our comprehensive experiments confirm the effectiveness of this counterfactual inference-based data augmentation for DSAs prediction.

Drug repositioning is the process of identifying potential associations between approved drugs and diseases (DDAs) to unveil novel therapeutic applications. Unlike traditional drug discovery approaches, a key advantage of drug repositioning lies in its capacity to leverage the existing knowledge and safety profiles of established medications, leading to significant reductions in both the time and costs associated with drug development. While various methods have been proposed to address this challenge using diverse strategies, the conventional approach for training DDAs prediction models typically relies on random sampling of unknown drug-disease pairs to construct negative samples. However, this method may inadvertently introduce unwanted noise or errors by erroneously categorizing some genuine DDAs as negative samples, thereby leaving room for improvement in current methodologies. In this paper, we introduce a novel negative sample selection algorithm for DDAs prediction that explicitly incorporates causal knowledge inherent in DDAs. To accomplish this, we first construct a heterogeneous information network (HIN) that encompasses various biological entities associated with DDAs and their interconnections. Subsequently, we utilize the outcomes of community detection within the HIN as a form of counterfactual inference, resulting in the development of a negative sample selection algorithm based on a thoughtfully designed counterfactual question. By combining the known DDAs (i.e., positive samples) with the newly generated negative samples, we train a prediction model that incorporates a graph learning module to acquire representations of drugs and diseases. Comprehensive experiments confirm the effectiveness of our proposed model for DDAs prediction.

Drug-drug interaction (DDI) pertains to the occurrence where the concomitant use of two or more drugs may lead to interactions in terms of their pharmacokinetic or pharmacodynamic behavior, resulting in unexpected effects. Accurately predicting DDIs holds significant importance in ensuring drug safety. Despite the numerous approaches proposed for DDI prediction, a majority of these methods often overlook the challenge presented by cold-start scenario, consequently limiting their applicability. This paper presents a novel data augmentation approach for the prediction of DDIs in cold-start scenarios. This method leverages counterfactual inference to generate meaningful pseudo samples for drugs with limited prior information. To achieve this, a HIN relevant to DDIs is initially established by amalgamating various associations between drugs and proteins. Subsequently, the identification of drug communities within this HIN is regarded as a form of counterfactual inference treatment, facilitating the generation of counterfactual links for cold-start drugs and thereby augmenting the training dataset. Lastly, we enhance our understanding of drug characteristics through a meta-path-based fusion mechanism, ultimately improving the accuracy of DDIs prediction in cold-start scenarios. We substantiate the effectiveness of our proposed method through an extensive series of experiments.

Microbe-drug interactions, which refer to the effects of drugs on microorganisms, play a crucial role in the realm of studying antibiotic-resistant bacteria and the development of antimicrobial agents. With the rapid progress in biomedical field, numerous experimental results containing validated microbe-drug interactions have been available in scientific articles. However, since failing to employ domain knowledge, traditional natural language processing methods encounter challenges in accurately identifying microbe and drug entities. Moreover, the unstructured characteristics and semantic complexity of biomedical literature pose difficulties for conventional text mining approaches to accurately grasp the syntactic features. In this paper, we present a novel microbial-drug relation extraction model called D-GCN, in which dual graph convolutional networks are used. Specifically, the drug database Drugbank is leveraged as external domain knowledge, while the graph convolutional network-based model SemGCN is utilized to learn meaningful features from biomedical texts. In addition, the attention graph convolutional network A-GCN is introduced to capture crucial syntactic features contained in texts. The experimental results show that the proposed model achieves better performance over the selected baseline models, which means D-GCN can not only accurately recognize microbial and drug entity representations, but also effectively identify the microbe-drug interactions.

Drug-drug interactions (DDIs) are compound effects when patients take two or more drugs at the same time, which may weaken the efficacy of drugs or cause unexpected side effects. Thus, accurately predicting DDIs is of great significance for the drug development and the drug safety surveillance. Although many methods have been proposed for the task, the biological knowledge related to DDIs is not fully utilized and the complex semantics among drug-related biological entities are not effectively captured in existing methods, leading to suboptimal performance. Moreover, the lack of interpretability for the predicted results also limits the wide application of existing methods for DDIs prediction. In this study, we propose a novel framework for predicting DDIs with interpretability. Specifically, we construct a heterogeneous information network (HIN) by explicitly utilizing the biological knowledge related to the procedure of inducing DDIs. To capture the complex semantics in HIN, a meta-path-based information fusion mechanism is proposed to learn high-quality representations of drugs. In addition, an attention mechanism is designed to combine semantic information obtained from meta-paths with different lengths to obtain final representations of drugs for DDIs prediction. Comprehensive experiments are conducted on 2410 approved drugs, and the results of predictive performance comparison show that our proposed framework outperforms selected representative baselines on the task of DDIs prediction. The results of ablation study and cold-start scenario indicate that the meta-path-based information fusion mechanism red is beneficial for capturing the complex semantics among drug-related biological entities. Moreover, the results of case study demonstrate that the designed attention mechanism is able to provide partial interpretability for the predicted DDIs. Therefore, the proposed method will be a feasible solution to the task of predicting DDIs.

Side effects of drugs have gained increasing attention in the biomedical field, and accurate identification of drug side effects is essential for drug development and drug safety surveillance. Although the traditional pharmacological experiments can accurately detect the side effects of drugs, the identifying process is time-consuming, costly, and may lead to incomplete identification of side effects. With the expanding of various biomedical databases, many computational methods have been developed for the task of drug-side effect associations (DSAs) prediction. However, existing methods have the following three drawbacks: 1). multiple drug-related databases are not fully used; 2). the complex semantics among drugs and side effects are not effectively captured; 3). the explainability of the predicted DSAs is missed for most existing methods. Therefore, there is an urgent need to find a more effective method for predicting DSAs. To address these issues, we propose a novel meta-path-based graph neural network model for drug-side effect associations prediction (MPGNN-DSA). In MPGNN-DSA, a heterogeneous information network is first constructed by combining multiple biological datasets. Then, a meta-path-based feature learning module is utilized for learning high-quality representations of drugs and side effects by capturing the semantics contained in meta-paths of the constructed HIN. With the learned features, the prediction module is conducted to derive the predicted side effects for drugs. In addition, the explainability of the predicted DSAs can be provided as well with the semantics contained in meta-paths. We conduct comprehensive experiments, and the results demonstrate the effectiveness of MPGNN-DSA, suggesting that the proposed method will be a feasible solution to the task of DSAs prediction.

In the treatment of bacterial infectious diseases, overuse of antibiotics may lead to not only bacterial resistance to antibiotics but also dysbiosis of beneficial bacteria which are essential for maintaining normal human life activities. Instead, phage therapy, which invades and lyses specific pathogenic bacteria without affecting beneficial bacteria, becomes more and more popular to treat bacterial infectious diseases. For the effective phage therapy, it requires to accurately predict potential phage-host interactions from heterogeneous information network consisting of bacteria and phages. Although many models have been proposed for predicting phage-host interactions, most methods fail to consider fully the sparsity and unconnectedness of phage-host heterogeneous information network, deriving the undesirable performance on phage-host interactions prediction. To address the challenge, we propose an effective model called GERMAN-PHI for predicting Phage-Host Interactions via Graph Embedding Representation learning with Multi-head Attention mechaNism. In GERMAN-PHI, the multi-head attention mechanism is utilized to learn representations of phages and hosts from multiple perspectives of phage-host associations, addressing the sparsity and unconnectedness in phage-host heterogeneous information network. More specifically, a module of GAT with talking-heads is employed to learn representations of phages and bacteria, on which neural induction matrix completion is conducted to reconstruct the phage-host association matrix. Results of comprehensive experiments demonstrate that GERMAN-PHI performs better than the state-of-the-art methods on phage-host interactions prediction. In addition, results of case study for two high-risk human pathogens show that GERMAN-PHI can predict validated phages with high accuracy, and some potential or new associated phages are provided as well.

Inferring microbe-metabolite interactions is conducive to understand how microbes affect human health, and specific microbial metabolites can be used as biomarkers for the diagnosis and treatment. Most of the existing methods only utilize the known mechanisms between microbiome and metabolome, while ignore the intragroup biological interactions(including microbe-microbe correlations and metabolite-metabolite correlations). In this paper, we propose a microbe-metabolite heterogeneous network fusion model based on graph convolution network (MMHNF) for inferring microbe-metabolite interactions. The proposed model not only utilizes the common properties of microbe-metabolite interactions, but also applies the properties of microbe-microbe and metabolite-metabolite correlation networks. In addition, it can learn low-dimensional and effective feature representations from multisource heterogeneous networks by applying the graph convolution network. Comprehensive experiments are performed on both simulated and experimental data, and the results show that MMHNF outperforms selected baselines. Moreover, results of case studies on inflammatory bowel disease (IBD) demonstrate further the effectiveness of the proposed model.

Microbes are closely related to human diseases, and the internal interactions among microbes influence disease production and development as well. Therefore, accurately predicting the associations between microbes and diseases becomes very important. This field is being carried out by a large number of researchers. However, existing methods have two deficiencies: (i) they fail to utilize the pathogenic impact of microbial interactions; (ii) they ignore the noise effect of unknown microbe-disease associations for training the prediction model. In this paper, we propose a novel model for above the shortcomings. Results of extensive experiments show that the proposed model achieves better performance over SOTA methods. In addition, the results of a case study on Type 1 diabetes demonstrate the effectiveness of the predicted associations between microbes and diseases derived from our model.

Drug side effect is an important entity in the biomedical field, and identifying the association of the drug-side effects is a very important issue in pharmacological studies and drug risk-benefit. Traditional side effect discovery methods are mainly based on pharmacological experiments. These methods can detect the side effects of some drugs, but the identification process is time-consuming, expensive, and fails to identify some rare side effects. In recent years, with the expansion of massive biomedical data, computational-based methods are widely developed and applied for the task of drug-side effect association(DSA) prediction. However, existing methods cannot fully utilize public biomedical databases, and the complex semantic associations between drugs and side effects are not effectively captured, which leads to suboptimal model prediction performance. In this study, we develop a novel meta-path-based graph neural network model for drug-side effect association prediction. In the proposed model, we first construct a heterogeneous information network(HIN) by fusing multiple biological datasets. And then, a novel meta-path-based feature learning module is designed to learn high-quality representations of drugs and side effects. Finally, with the learned features, the prediction module utilizes a fully connected neural network to make prediction. In addition, comprehensive experiments is conducted, the results demonstrate the effectiveness of our model, indicating that the method will be a viable approach for DSA prediction tasks.

Compared with traditional methods, drug repositioning is a viable solution to drug discovery. Drug repositioning usually applies the procedure of drug-disease associations (DDAs) prediction, which can reduce the cost and time of drug development and improve the success rate of drug discovery. In this paper, we develop a new multi-task learning framework based on heterogeneous graph convolutional network (MTHGCN) to recognize potential DDAs. In MTHGCN, a heterogeneous information network is constructed by combining multiple biological datasets. And then, a module based on graph convolutional networks is utilized to learn low-dimensional representations of drugs and diseases. Finally, we design two types of auxiliary tasks to help to train the target DDAs prediction task based on the multi-task learning mechanism. We conduct comprehensive experiments on MTHGCN. The results demonstrate the effectiveness of MTHGCN for drug repositioning.

Drug-drug interactions (DDIs) refer to the compound effects that may impair the effectiveness of drugs or cause unexpected side effects when two or more drugs are taken together. Therefore, it is very important to accurately predict DDIs for the drug development and drug safety monitoring. Many methods have been proposed to accomplish this task, but the existing methods fail to make full use of the biological knowledge related to DDIs, and do not effectively capture the complex semantics between biological entities related to drugs, resulting in poor performance. In this paper, we propose a novel DDIs prediction framework based on heterogeneous information network (HIN). More specifically, we construct the HIN which combines biological knowledge related to DDIs. In order to capture the complex semantics in HIN, a meta-path-based fusion mechanism is proposed to obtain high-quality drugs’ representations. Moreover, we design a meta-path level attention to combine semantics obtained from meta-paths with different lengths to obtain the final representations of drugs for DDIs prediction. The experimental results demonstrate that the framework performs better than the selected representative baselines on 2410 approved drugs.

The increasing prevalence of antibiotic resistance has become a global health crisis. For the purpose of safety regulation, it is of high importance to identify antibiotic resistance genes (ARGs) in bacteria. Although culture-based methods can identify ARGs relatively more accurately, the identifying process is time-consuming and specialized knowledge is required. With the rapid development of whole genome sequencing technology, researchers attempt to identify ARGs by computing sequence similarity from public databases. However, these computational methods might fail to detect ARGs due to the low sequence identity to known ARGs. Moreover, existing methods cannot effectively address the issue of multidrug resistance prediction for ARGs, which is a great challenge to clinical treatments. To address the challenges, we propose an end-to-end multi-label learning framework for predicting ARGs. More specifically, the task of ARGs prediction is modeled as a problem of multi-label learning, and a deep neural network-based end-to-end framework is proposed, in which a specific loss function is introduced to employ the advantage of multi-label learning for ARGs prediction. In addition, a dual-view modeling mechanism is employed to make full use of the semantic associations among two views of ARGs, i.e. sequence-based information and structure-based information. Extensive experiments are conducted on publicly available data, and experimental results demonstrate the effectiveness of the proposed framework on the task of ARGs prediction.

Over the past decades, Chemical-induced Disease (CID) relations have attracted extensive attention in biomedical community, reflecting wide applications in biomedical research and healthcare field. However, prior efforts fail to make full use of the interaction between local and global contexts in biomedical document, and the derived performance needs to be improved accordingly. In this paper, we propose a novel framework for document-level CID relation extraction. More specifically, a stacked Hypergraph Aggregation Neural Network (HANN) layers are introduced to model the complicated interaction between local and global contexts, based on which better contextualized representations are obtained for CID relation extraction. In addition, the CID Relation Heterogeneous Graph is constructed to capture the information with different granularities and improve further the performance of CID relation classification. Experiments on a real-world dataset demonstrate the effectiveness of the proposed framework.

Prescription opioids are powerful pain-reducing medications. Thousands of articles that focus on prescription opioid use (POU) and its associated medical disorders have been published. However, it is time-consuming and labor-intensive to extract and understand the information of all POU-related published articles. In this study, we applied the well-adapted topic modeling method, Latent Dirichlet Allocation (LDA), to perform text mining on POU-related literature. We have collected six large academic abstract datasets by searching PubMed using the Medical Subject Headings (MeSH): prescription opioid, codeine, morphine, hydrocodone, oxycodone, and methadone. We then applied topic modeling to identify topics and analyze topic similarities/differences in these six datasets. Word clouds and histograms were used to depict the distribution of vocabularies over each topic in which the most prevalent words conveyed a topic’s meaning. TreeMap and trend analysis were performed to fingerprint abstracts and explore the prevalent topic dynamics in the POU-related literature. Results showed the ability of topic modeling as a computational tool to segregate a vast quantity of articles into different themes that provide a systematic literature overview. The LDA topics recaptured the search keywords in PubMed and revealed further relevant themes by comparison analysis between different datasets.

With the increasing resistance of bacteria to antibiotics, the problem of antibiotic resistance has become a major challenge in healthcare. Therefore, the accurate identification of antibiotic resistance genes (ARGs) in bacteria is particularly urgent and important. Although culture-based methods can accurately identify ARGs, they are limited by domain knowledge and are time-consuming. With the availability of whole genome sequence data, most existing methods rely on sequence alignment to calculate sequence similarity to predict ARGs. However, for ARGs that has low sequence identity with known sequences, the existing methods do not identify them well. To address the challenges, we propose a multiple view learning based framework for ARGs. Experimental results demonstrate that the proposed approach performs better than the selected representative baselines on the real-world dataset.

Objective: To classify causal associations among drugs and adverse events by the identified drug safety signals from the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS).Material and Methods: FAERS reports were collected for a period between 2004 and 2014. Empirical Bayes Geometric Mean was applied to model associations between drugs and adverse events. Identified signals were evaluated using Reporting Ratio values and Chi-Square test. Based on the identified drug-associated adverse events, we constructed a drug-drug association network, and applied a random walk algorithm to find drug communities with similar adverse event patterns. We developed 14 clusters for comparison with the 14 main groups in the first level of the Anatomical Therapeutic Chemical (ATC) classification system to evaluate relationships between the two classification systems.Results: The retrieved FAERS dataset included 981 drugs and 16,179 adverse events, from which we identified 63,083 significant drug-adverse event pairs. We found new potential safety signals when comparing the drug-adverse event pairs with information from relevant sources. Network analysis of the constructed drug communities revealed connections among drugs, adverse events, and ATC codes, suggesting that drug adverse events may be used to predict the ATC codes of unclassified drugs. For the ATC-classified drugs, the network analysis revealed potential relationships among the drugs by calculating the similarities of the adverse events.Conclusion: The generated drug groups and the drug-drug associations derived from the network analysis might have predictive potential for adverse events, as well as provide information for drug review and development.

The main goal of biomedical event extraction is to structurally extract biomedical events from texts, however, the specificity of the domain makes both text modeling and data annotation very difficult. We propose a self-supervised learning-based data augmentation method in this paper and design specific augmentation strategies for biomedical entities and event triggers in biomedical texts, which solves the problem of sparse annotation data to some extent. In addition we improve the reinforcement learning-based event extraction method to improve the training efficiency of the model. The experiments on two datasets demonstrate the effectiveness of our method.

MOTIVATION: Multiple events extraction from biomedical literature is a challenging task for biomedical community. Usually, biomedical event extraction is modeled as two sub-tasks, trigger identification and argument detection. Most existing methods perform these two sub-tasks sequentially, and fail to make full use of the interaction between them, leading to suboptimal results for multiple biomedical events extraction. RESULTS: We propose a novel framework of reinforcement learning (RL) for the task of multiple biomedical events extraction. More specifically, trigger identification and argument detection are treated as main-task and subsidiary-task, respectively. Assigning the event type of triggers (in the main-task) is viewed as the action taken in RL, and the result of corresponding argument detection (i.e. the subsidiary-task) for the identified trigger is used for computing the reward of the taken action. Moreover, the result of the subsidiary-task is modeled as part of environment information in RL to help the procedure of trigger identification. In addition, external biomedical knowledge bases are employed for representation learning of biomedical text, which can improve the performance of biomedical event extraction. Results on two widely used biomedical corpora demonstrate that the proposed framework performs better than the selected baselines on the task of multiple events extraction. The ablation test indicates the contributions of RL and external KBs to the performance improvement in the proposed method. In addition, by modeling multiple events extraction under the RL framework, the supervised information is exploited more effectively than the classical supervised learning paradigm. Availability and implementationSource codes will be available at: https://github.com/David-WZhao/BioEE-RL.

Recent years have witnessed the rapidly growing of the amount of semistructured documents in real-world applications. Due to the huge size of the real-world data, how to manage semistructured documents effectively is a big challenge for researchers. As a fundamental task in natural language processing field, document classification is a feasible way to handle the large-scale semistructured documents. However, existing methods fail to explicitly take advantage of the hierarchical semantics in semistructured documents. It's known that the contained semantics is beneficial for understanding the semistructured documents. Considering the hierarchical structure of a given semistructured document, we propose a semistructured document classification framework which explicitly utilizes the semantic hierarchical attention mechanism. More specifically, the hierarchical attention mechanism and graph neural network are employed to model semistructured documents, by which the multilevel semantic relationships and grammatical information are considered. Moreover, we propose an adaptive class cost learning method to treat the issue of data imbalance. Comprehensive experiments are conducted on two real-world data sets, and the results demonstrate that our framework performs better than selected baselines for semistructured document classification.

To learn a well-performed image annotation model, a large number of labeled samples are usually required. Although the unlabeled samples are readily available and abundant, it is a difficult task for humans to annotate large numbers of images manually. In this article, we propose a novel semi-supervised approach based on adaptive weighted fusion for automatic image annotation that can simultaneously utilize the labeled data and unlabeled data to improve the annotation performance. At first, two different classifiers, constructed based on support vector machine and covolutional neural network, respectively, are trained by different features extracted from the labeled data. Therefore, these two classifiers are independently represented as different feature views. Then, the corresponding features of unlabeled images are extracted and input into these two classifiers, and the semantic annotation of images can be obtained respectively. At the same time, the confidence of corresponding image annotation can be measured by an adaptive weighted fusion strategy. After that, the images and its semantic annotations with high confidence are submitted to the classifiers for retraining until a certain stop condition is reached. As a result, we can obtain a strong classifier that can make full use of unlabeled data. Finally, we conduct experiments on four datasets, namely, Corel 5K, IAPR TC12, ESP Game, and NUS-WIDE. In addition, we measure the performance of our approach with standard criteria, including precision, recall, F-measure, N+, and mAP. The experimental results show that our approach has superior performance and outperforms many state-of-the-art approaches.

Biomedical event extraction aims to detect fine-grained interactions between biomedical entities in biomedical texts, and has become a research hotspot for researchers in the biomedical field. As the first step in biomedical event extraction, biomedical event trigger identification plays an important role in the whole process. Although researchers have proposed many methods, the performance of existing methods is not desirable due to the reliance on a large number of labeled training samples and the need for expert knowledge in the biomedical field. To treat this issue, we model the biomedical event trigger identification as a few-shot learning problem. Specifically, we utilize a knowledge-aware attention layer to obtain a rich informative representation for entities, and improve the derived prototypes accordingly by prototypical network. In addition, the module of relation network is introduced to train a more reasonable distance function for trigger type prediction. The results demonstrate the effectiveness of the proposed framework according to F1-score.

In recent years, Chemical-induced Disease (CID) relations are the most searched topics by PubMed users worldwide, reflecting its extensive applications in biomedical research and public health field. However, for CID relation extraction, prior methods fail to make full use of the interaction between local and global contexts in biomedical document. To better capture the complex relationships among contexts, we propose an effective framework for document-level CID relation extraction. Specifically, the stacked Hypergraph Aggregation Neural Network (HANN) layers are applied to model effectively the interaction between local and global contexts. Moreover, by constructing CID Relation Heterogeneous Graph, we can capture the different granularities of information and learn better contextualized representations for CID relation extraction. Extensive experiments on a commonly used dataset demonstrate the effectiveness of the proposed method.

Biomedical event extraction is usually modeled as two sub-tasks: trigger identification and argument detection. Most existing methods perform these two sub-tasks sequentially but ignore the interaction between them. This paper proposes a novel method for multiple biomedical events extraction, in which the task of event extraction is modeled under a framework of reinforcement learning (RL). We treat the trigger identification and argument detection as main-task and subsidiary-task, respectively. And the result of argument detection is modeled as environmental information. In this way, the proposed method can capture the interaction between two sub-tasks, and the semantic associations among multiple biomedical events are also utilized effectively. Moreover, external biomedical knowledge bases are employed for representation learning of biomedical text. Comprehensive experiments are conducted on two widely used biomedical corpora, and results demonstrate that our method gains better performance compared to existing methods, especially in multiple biomedical events extraction.

Biomedical event extraction has wide applications in biomedicine field. As a prerequisite step in biomedical event extraction, event trigger identification has attracted growing attention in biomedical research. Although many approaches have been proposed for biomedical event trigger identification, two main challenges still remain for researchers: 1) most of the existing approaches treat each sentence separately in biomedical documents, failing to make full use of the semantics in the global document context; 2) the sparseness of event triggers leads to a serious issue of imbalanced class for trigger identification. In this paper, we propose an end-to-end framework for biomedical event trigger identification which addresses effectively the two mentioned challenges accordingly. Specifically, a hierarchical attention mechanism is used to model the global document context, including the semantic relationships both among words in the same sentence and among sentences in the same document. In addition, an adaptive class weight learning method is proposed to treat the class imbalance issue in biomedical event trigger identification. Experimental results on two commonly used datasets demonstrate the effectiveness of the proposed framework.

Biomedical document classification is a fundamental task in biomedical field. Existing methods do not make full use of the hierarchically semantic structures in biomedical documents which can be utilized to improve the performance of biomedical document classification. In this paper, according to the hierarchical structures in given biomedical documents, we propose two models for biomedical document classification, which are based on the semantically hierarchical attention mechanism. Specifically, we utilize a hierarchical attention mechanism to model biomedical documents, taking into account simultaneously multiple-level semantic relationships in documents. In addition, an adaptive cost sensitive learning method is proposed to address the data imbalance issue. Extensive experiments on two real-world datasets demonstrate the effectiveness of the proposed methods.

To learn a well-performed image annotation model, a large number of labeled samples are usually required. Although the unlabeled samples are readily available and abundant, it's a difficult task for humans to annotate large amounts of images manually. In this paper, we propose a novel semi-supervised approach based on co-training algorithm for automatic image annotation, which can utilize the labeled data and unlabeled data for the system simultaneously. Firstly, two different classifiers, namely the CNN (convolutional neural network) and the LDA-SVM, are constructed by all the labeled data. These two classifiers are independently represented as different feature views. Then, the most confident data with relevant pseudo-labels are chosen and amalgamated with the whole labeled dataset. After that, the two classifiers are retrained with the new labeled dataset until a stop condition is reached. In each iteration process, the unlabeled samples are labeled by high confidence pseudo-labels that are estimated by an adaptive weighted fusion method. Finally, we conduct experiments on two datasets, namely, IAPR TC-2 and NUS-WIDE, and measure the performance of the model with standard criteria, including precision, recall, F-measure, N+ and mAP. The experimental results show that our approach has superior annotation performance and outperforms many state-of-the-art automatic image annotation approaches.

To learn a well-performed image annotation model, a large number of labeled samples are usually required. In this paper, we propose a novel semi-supervised approach based on adaptive weighted fusion for automatic image annotation, which can utilize the labeled data and unlabeled data simultaneously. Firstly, two different classifiers, namely the CNN (convolutional neural network) and the LDA-SVM, are constructed by all the labeled data. These two classifiers are independently represented as different feature views. Then, the most confident data with relevant pseudo-labels are chosen and amalgamated with the whole labeled dataset. After that, the two classifiers are retrained with the new labeled dataset until a stop condition is reached. In each iteration process, the unlabeled samples are labeled by high confidence pseudo-labels that are estimated by an adaptive weighted fusion strategy. Finally, we conduct experiments on two datasets, namely IAPR TC12 and NUS-WIDE, and measure the performance of the model with standard criteria, including precision, recall, F-measure, N+ and mAP. The experimental results show that our approach outperforms many state-of-the-art automatic image annotation approaches.

Recommending valuable contents for microblog users is an important way to improve users' experiences. As high quality descriptors of user semantics, tags have always been used to represent users' interests or attributes. In this work, we propose a microblog recommendation approach via hypergraph random walk tag expansion and user social relation. More specifically, microblogs are considered as hyperedges and terms are taken as hypervertexs for each user, and the weighting strategies for both hyperedges and hypervertexs are established. Random walk is performed on the weighted hypergraph to obtain a number of terms as tags for users. And then the tag similarity matrix and the user-tag matrix can be constructed based on tag probability correlations and weight of each tag. Besides, the significance of user social relation is also considered for recommendation. Moreover, an iterative updating scheme is developed to get the final user-tag matrix for computing the similarities between microblogs and users. Experimental results show that the algorithm is effective in microblog recommendation.

Network clustering is an essential approach to finding latent clusters in real-world networks. As the scale of real-world networks becomes increasingly larger, the existing network clustering algorithms fail to discover meaningful clusters efficiently. In this paper, we propose a framework called AnySCAN, which applies anytime theory to the structural clustering algorithm for networks (SCAN). Moreover, an active learning strategy is proposed to advance the refining procedure in AnySCAN framework. AnySCAN with the active learning strategy is able to find the exactly same clustering result on large-scale networks as the original SCAN in a significantly more efficient manner. Extensive experiments on real-world and synthetic networks demonstrate that our proposed method outperforms existing network clustering approaches.

BACKGROUND: Next-generation sequencing (NGS) technologies have provided researchers with vast possibilities in various biological and biomedical research areas. Efficient data mining strategies are in high demand for large scale comparative and evolutional studies to be performed on the large amounts of data derived from NGS projects. Topic modeling is an active research field in machine learning and has been mainly used as an analytical tool to structure large textual corpora for data mining. METHODS: We report a novel procedure to analyse NGS data using topic modeling. It consists of four major procedures: NGS data retrieval, preprocessing, topic modeling, and data mining using Latent Dirichlet Allocation (LDA) topic outputs. The NGS data set of the Salmonella enterica strains were used as a case study to show the workflow of this procedure. The perplexity measurement of the topic numbers and the convergence efficiencies of Gibbs sampling were calculated and discussed for achieving the best result from the proposed procedure. RESULTS: The output topics by LDA algorithms could be treated as features of Salmonella strains to accurately describe the genetic diversity of fliC gene in various serotypes. The results of a two-way hierarchical clustering and data matrix analysis on LDA-derived matrices successfully classified Salmonella serotypes based on the NGS data. The implementation of topic modeling in NGS data analysis procedure provides a new way to elucidate genetic information from NGS data, and identify the gene-phenotype relationships and biomarkers, especially in the era of biological and medical big data. CONCLUSION: The implementation of topic modeling in NGS data analysis provides a new way to elucidate genetic information from NGS data, and identify the gene-phenotype relationships and biomarkers, especially in the era of biological and medical big data.

After publication of the original article [1] it was brought to our attention that the following was incorrectly placed under subheading '3.Classification analysis and comparison' of subsection 'Evaluation of topic modeling performance' of the 'Methods' section:Topic model-derived clustering method [33] was applied, in which LDA was utilized as a feature reduction approach for cluster analysis.The LDAderived topics were considered as the new features of datasets.The sampletopic matrix (Fig. 1(f)) was treated as a new representation of the original dataset.Based on the sample-topic matrix (topic number was chosen as 5 and 30, respectively), conventional clustering algorithms, such as k-means, was used for the clustering analysis.The number of clusters was set as 7 in the k-means method due to 7 different serotypes in the dataset.While in comparison, k-means algorithm was also applied on VSM matrix using Hamming Distance similarities.For further comparison, due to the dimension reduction of topic modeling approach, the traditional tool of PCA was used to reduce features (Numbers of 2, 5, 10 and 30 were randomly selected as the reduced features, respectively) of VSM matrix followed by the k-means cluster analysis.Moreover, clustering by only LDA referred as "highest probable topic assignment" [33] (5 and 30 topics were used) was also used for comparison.In "highest probable topic assignment", the LDA-derived topics were made as the clusters of the dataset.Then, each sample was assigned to the cluster (Topic) with the highest probability in the row of the sample-topic matrix.To interpret the clustering results obtained by the k-means algorithm, samples in each cluster were labeled as the dominant serotype of the samples in the cluster.

In this paper, we present a semi-supervised clustering method for microblog in which both word-level and microblog (document)-level constraints are automatically generated totally based on statistical information rather than any kind of external knowledge. The key idea is first to explore term correlation data, which investigates both inter and intra correlation of words, and the initial similarity between words can therefore be deduced. And then an iterative method is established to calculate both word similarity and microblog similarity. The mechanism of incorporating dual constraints is presented based on word similarity and microblog similarity. We then formulate short text clustering problem as a non-negative matrix factorization based on dual constraints. Empirical study of two real-world dataset shows the superior performance of our framework in handling noisy and microblogs.

BACKGROUND: Topic modelling is an active research field in machine learning. While mainly used to build models from unstructured textual data, it offers an effective means of data mining where samples represent documents, and different biological endpoints or omics data represent words. Latent Dirichlet Allocation (LDA) is the most commonly used topic modelling method across a wide number of technical fields. However, model development can be arduous and tedious, and requires burdensome and systematic sensitivity studies in order to find the best set of model parameters. Often, time-consuming subjective evaluations are needed to compare models. Currently, research has yielded no easy way to choose the proper number of topics in a model beyond a major iterative approach. METHODS AND RESULTS: Based on analysis of variation of statistical perplexity during topic modelling, a heuristic approach is proposed in this study to estimate the most appropriate number of topics. Specifically, the rate of perplexity change (RPC) as a function of numbers of topics is proposed as a suitable selector. We test the stability and effectiveness of the proposed method for three markedly different types of grounded-truth datasets: Salmonella next generation sequencing, pharmacological side effects, and textual abstracts on computational biology and bioinformatics (TCBB) from PubMed. CONCLUSION: The proposed RPC-based method is demonstrated to choose the best number of topics in three numerical experiments of widely different data types, and for databases of very different sizes. The work required was markedly less arduous than if full systematic sensitivity studies had been carried out with number of topics as a parameter. We understand that additional investigation is needed to substantiate the method's theoretical basis, and to establish its generalizability in terms of dataset characteristics.

BACKGROUND: The big data moniker is nowhere better deserved than to describe the ever-increasing prodigiousness and complexity of biological and medical datasets. New methods are needed to generate and test hypotheses, foster biological interpretation, and build validated predictors. Although multivariate techniques such as cluster analysis may allow researchers to identify groups, or clusters, of related variables, the accuracies and effectiveness of traditional clustering methods diminish for large and hyper dimensional datasets. Topic modeling is an active research field in machine learning and has been mainly used as an analytical tool to structure large textual corpora for data mining. Its ability to reduce high dimensionality to a small number of latent variables makes it suitable as a means for clustering or overcoming clustering difficulties in large biological and medical datasets. RESULTS: In this study, three topic model-derived clustering methods, highest probable topic assignment, feature selection and feature extraction, are proposed and tested on the cluster analysis of three large datasets: Salmonella pulsed-field gel electrophoresis (PFGE) dataset, lung cancer dataset, and breast cancer dataset, which represent various types of large biological or medical datasets. All three various methods are shown to improve the efficacy/effectiveness of clustering results on the three datasets in comparison to traditional methods. A preferable cluster analysis method emerged for each of the three datasets on the basis of replicating known biological truths. CONCLUSION: Topic modeling could be advantageously applied to the large datasets of biological or medical research. The three proposed topic model-derived clustering methods, highest probable topic assignment, feature selection and feature extraction, yield clustering improvements for the three different data types. Clusters more efficaciously represent truthful groupings and subgroupings in the data than traditional methods, suggesting that topic model-based methods could provide an analytic advancement in the analysis of large biological or medical datasets.

The big data analytics community has accepted MapReduce as a programming model for processing massive data on distributed systems such as a Hadoop cluster. MapReduce has been evolving to improve its performance. We identified skewed workload among workers in the MapReduce ecosystem. The problem of skewed workload is of serious concern for massive data processing. We tackled the workload balancing issue by introducing a hierarchical MapReduce, or h-MapReduce for short. h-MapReduce identifies a heavy task by a properly defined cost function. The heavy task is divided into child tasks that are distributed among available workers as a new job in MapReduce framework. The invocation of new jobs from a task poses several challenges that are addressed by h-MapReduce. Our experiments on h-MapReduce proved the performance gain over standard MapReduce for data-intensive algorithms. More specifically, the increase of the performance gain is exponential in terms of the size of the networks. In addition to the exponential performance gains, our investigations also found a negative effect of deploying h-MapReduce due to an inappropriate definition of heavy tasks, which provides us a guideline for an effective application of h-MapReduce.

Big data such as complex networks with over millions of vertices and edges is infeasible to process using conventional computation. MapReduce is a programming model that empowers us to analyze big data in a cluster of computers. In this paper we propose a Parallel Structural Clustering Algorithm for big Networks (PSCAN) in MapReduce for the detection of clusters or community structures in big networks such as Twitter. PSCAN is based on the structural clustering algorithm of SCAN, which not only finds cluster accurately, but also identifies vertices playing special roles such as hubs and outliers. An empirical evaluation of PSCAN using both real and synthetic networks demonstrated an outstanding performance in terms of accuracy and running time. We analyzed a Twitter network with over 40 million users and 1.4 billion follower/following relationships by using PSCAN on a Hadoop cluster with 15 computers. The result shows that PSCAN successfully detected interesting communities of people who share common interests.

Substantial percent of global Internet users are now actively use Twitter. In recent times, Twitter has been experiencing explosive growth, attracting celebrities consequently a growing mass of user coverage. Insights of such a social network aid researchers in understanding behavioral dynamics of the society. Though there have been attempts to study social networks, they did not scale to process social networks on the scale of Twitter user-follower network. In this paper we uncovered some of the essential properties of the complete Twitter user-follower network. The properties include degree distribution, connectivity, strength of following relationships, clustering coefficient. Our investigations showed that the Twitter user-follower network follows power-law degree distribution. We found Twitter being a connected network. The strength of the relationships among users is distributed nearly uniform on the scale of 0.0 to 1.0. Nearly 90% of the users possess '0' clustering coefficient, which refers to the least possibility to find communities in the network. In addition to the listed properties, this study found communities of users with high clustering coefficient despite many users with low clustering coefficient. A sample of the communities is validated manually for accuracy. The validation proved that the communities are representing users of similar interests. The communities found from this work yields to friend recommendations, target based advertisements, etc.

BACKGROUND: Pulsed field gel electrophoresis (PFGE) is currently the most widely and routinely used method by the Centers for Disease Control and Prevention (CDC) and state health labs in the United States for Salmonella surveillance and outbreak tracking. Major drawbacks of commercially available PFGE analysis programs have been their difficulty in dealing with large datasets and the limited availability of analysis tools. There exists a need to develop new analytical tools for PFGE data mining in order to make full use of valuable data in large surveillance databases. RESULTS: In this study, a software package was developed consisting of five types of bioinformatics approaches exploring and implementing for the analysis and visualization of PFGE fingerprinting. The approaches include PFGE band standardization, Salmonella serotype prediction, hierarchical cluster analysis, distance matrix analysis and two-way hierarchical cluster analysis. PFGE band standardization makes it possible for cross-group large dataset analysis. The Salmonella serotype prediction approach allows users to predict serotypes of Salmonella isolates based on their PFGE patterns. The hierarchical cluster analysis approach could be used to clarify subtypes and phylogenetic relationships among groups of PFGE patterns. The distance matrix and two-way hierarchical cluster analysis tools allow users to directly visualize the similarities/dissimilarities of any two individual patterns and the inter- and intra-serotype relationships of two or more serotypes, and provide a summary of the overall relationships between user-selected serotypes as well as the distinguishable band markers of these serotypes. The functionalities of these tools were illustrated on PFGE fingerprinting data from PulseNet of CDC. CONCLUSIONS: The bioinformatics approaches included in the software package developed in this study were integrated with the PFGE database to enhance the data mining of PFGE fingerprints. Fast and accurate prediction makes it possible to elucidate Salmonella serotype information before conventional serological methods are pursued. The development of bioinformatics tools to distinguish the PFGE markers and serotype specific patterns will enhance PFGE data retrieval, interpretation and serotype identification and will likely accelerate source tracking to identify the Salmonella isolates implicated in foodborne diseases.

This paper presents a framework that actively selects informative documents pairs for semi-supervised document clustering. The semi-supervised document clustering algorithm is a Constrained DBSCAN (Cons-DBSCAN), which incorporates instance-level constraints to guide the clustering process in DBSCAN. By obtaining user feedbacks, our proposed active learning algorithm can get informative instance level constraints to aid clustering process. Experimental results show that Cons-DBSCAN with the proposed active learning approach can provide an appealing clustering performance.