MOTIVATION: The accurate and robust representation of drug molecule features, the prediction of drug-target biomacromolecule interactions, and the determination of physicochemical properties are crucial in drug development. However, these tasks remain challenging due to issues such as the limited generalizability of single-modal representations, the absence of multitask prediction frameworks, and weak adaptability in cold-start scenarios. RESULTS: In this study, we present DrugDL, a framework for comprehensive drug molecule representation and the prediction of multiple downstream tasks, including drug-target interactions, binding affinities, binding sites, physicochemical properties, toxicity, and drug-drug interactions. DrugDL jointly learns representations of the drug chemical space and the target protein biological space, while capturing multiscale interaction mechanisms between drug molecules and target proteins through the integration of cross-modal contrastive learning and single-modal feature enhancement algorithms. Specifically, DrugDL employs a multitask prediction framework to predict multiple properties of drug molecules. In practical applications, it consistently outperforms state-of-the-art methods, particularly in cold-start tasks. The framework has been successfully applied to high-throughput screening, the identification of inhibitors of SARS-CoV-2 and metabolic enzymes, and the prediction of cancer-targeted drugs. Experimental validations on EGFR and ALK targets further demonstrate its effectiveness as a precise drug discovery tool. By enabling accurate molecular representation and multi-property prediction, DrugDL provides end-to-end technical support for drug development, thereby significantly accelerating the drug discovery process. AVAILABILITY AND IMPLEMENTATION: The datasets and code are available at https://github.com/ZhangQi9910/DrugDL. The version of record is archived in Zenodo with the DOI: 10.5281/zenodo.20579718.

Anticancer peptides (ACPs) represent a promising class of therapeutic agents that selectively destroy cancer cells while sparing healthy tissues. Despite their potential, biological challenges including poor biochemical stability, limited tumor selectivity, and inefficient delivery mechanisms hinder their clinical translation. In parallel, the rapid expansion of peptide sequence data underscores the urgent need for accurate, scalable, and generalizable ACP prediction models. To address these limitations, we propose a robust hybrid deep learning framework, termed as Mamba-ACP-that integrates transformer-based Evolutionary Scale Modeling (ESM-2) embeddings, handcrafted features (AAindex, BLOSUM62), and a Mamba-based sequence modeling architecture. This approach captures both evolutionary and physicochemical properties of peptides to enhance prediction performance. The model was trained and validated by using two benchmark datasets Set 1 and Set 2, commonly used in peptide-based computational biology. Mamba-ACP achieves 87.59% accuracy and an AUC of 0.9356 on Set 1, and 96.69% accuracy and an AUC of 0.9922 on Set 2, beating state-of-the-art ACP predictors like ACP-CapsPred and GRDF by a significant margin. The Mamba-ACP framework processes a token-level fused representation obtained by concatenating the token-level ESM-2 embeddings with the PCA-reduced handcrafted residue descriptors at each sequence position. These results affirm the effectiveness of combining pre-trained transformer embeddings with handcrafted features and structured sequence modeling in improving ACP classification. Our findings position Mamba-ACP as a new benchmark in computational peptide discovery, offering strong generalizability, reduced false positives, and efficient performance. We further provide model-level explanations via gradient-based residue/token saliency, SHAP feature importance for AAindex/BLOSUM62 descriptors, motif-level enrichment, and saliency-guided residue mutation validation.

Comprehending biological reproduction and cellular metabolism is facilitated by the Enzyme Commission, which matches protein sequences to the biochemical reactions they catalyse through EC numbers. In recent years, several methods have been proposed for predicting enzyme function. However, these methods still encounter challenges. Firstly, traditional methods for manually designing enzyme features are complex and cumbersome, lacking an effective generalized method for embedding enzyme sequences. Secondly, the distribution gap between different enzymes is significant, which resulting in existing methods struggling to predict multilevel enzyme functions. Thirdly, traditional enzyme function prediction models only extract single view feature of enzyme, so there is still room for further improving the ability of these models to extract enzyme data. To address these challenges, a new multilevel enzyme function prediction model (SMENET) based on multi-view semantics is proposed. This method uses protein large language model to extract semantic information. Subsequently, this semantic information is fed into multiple information extraction network modules, followed by using Biologic Sematic Attention to integrate these views' information. Finally, a multi-view adaptive fusion network is designed to extract the best common representation between multiple semantic views. Extensive experiments were conducted on multiple datasets to validate the effectiveness of SMENET.

Log statements have become an integral part of modern software systems.Prior research efforts have focused on supporting the decisions of placing log statements, such as where/what to log.With the increasing adoption of Large Language Models (LLMs) for coderelated tasks such as code completion or generation, automated approaches for generating log statements have gained much momentum.However, the performance of these approaches still has a long way to go.This paper explores enhancing the performance of LLM-based solutions for automated log statement generation by post-training LLMs with a purpose-built dataset.Thus the primary contribution is a novel approach called AUCAD, which automatically constructs such a dataset with information extracting from log-related issues.Researchers have long noticed that a significant portion of the issues in the open-source community are related to log statements.However, distilling this portion of data requires manual efforts, which is labor-intensive and costly, rendering it impractical.Utilizing our approach, we automatically extract logrelated issues from 1,537 entries of log data across 88 projects and identify 808 code snippets (i.e., methods) with retrievable source code both before and after modification of each issue (including log statements) to construct a dataset.Each entry in the dataset consists of a data pair representing high-quality and problematic log statements, respectively.With this dataset, we proceed to post-train multiple LLMs (primarily from the Llama series) for automated * Corresponding author.

High-throughput sequencing methods have brought about a huge change in omics-based biomedical study. Integrating various omics data is possibly useful for identifying some correlations across data modalities, thus improving our understanding of the underlying biological mechanisms and complexity. Nevertheless, most existing graph-based feature extraction methods overlook the complementary information and correlations across modalities. Moreover, these methods tend to treat the features of each omics modality equally, which contradicts current biological principles. To solve these challenges, we introduce a novel approach for integrating multi-omics data termed Multi-Omics hypeRgraph integration nEtwork (MORE). MORE initially constructs a comprehensive hyperedge group by extensively investigating the informative correlations within and across modalities. Subsequently, the multi-omics hypergraph encoding module is employed to learn the enriched omics-specific information. Afterward, the multi-omics self-attention mechanism is then utilized to adaptatively aggregate valuable correlations across modalities for representation learning and making the final prediction. We assess MORE's performance on datasets characterized by message RNA (mRNA) expression, Deoxyribonucleic Acid (DNA) methylation, and microRNA (miRNA) expression for Alzheimer's disease, invasive breast carcinoma, and glioblastoma. The results from three classification tasks highlight the competitive advantage of MORE in contrast with current state-of-the-art (SOTA) methods. Moreover, the results also show that MORE has the capability to identify a greater variety of disease-related biomarkers compared to existing methods, highlighting its advantages in biomedical data mining and interpretation. Overall, MORE can be investigated as a valuable tool for facilitating multi-omics analysis and novel biomarker discovery. Our code and data can be publicly accessed at https://github.com/Wangyuhanxx/MORE.

Accurately identifying protein functions is essential to understand life mechanisms and thus advance drug discovery. Although biochemical experiments are the gold standard for determining protein functions, they are often time-consuming and labor-intensive. Here, we proposed a novel composite deep-learning method, Multi-source Knowledge Fusion for Gene Ontology prediction (MKFGO), to infer Gene Ontology (GO) attributes through integrating five complementary pipelines built on multi-source biological data. MKFGO was rigorously benchmarked on 1522 nonredundant proteins, demonstrating superior performance over 12 state-of-the-art function prediction methods. Comprehensive data analyses revealed that the major advantage of MKFGO lies in its two deep-learning components, handcrafted feature representation-based GO prediction (HFRGO) and protein large language model (PLM)-based GO prediction (PLMGO), which derive handcrafted features and PLM-based features, respectively, from protein sequences in different biological views, with effective knowledge fusion at the decision-level. HFRGO leverages a long short-term memory (LSTM)-attention network embedded with handcrafted features, in which the triplet loss-based guilt-by-association strategy is designed to enhance the correlation between feature similarity and function similarity. PLMGO employs the PLM to capture feature embeddings with discriminative functional patterns from sequences. Meanwhile, another three components provide complementary insights for further improving prediction accuracy, driven by protein-protein interaction, GO term probability, and protein-coding gene sequence, respectively. The source codes and models of MKFGO are freely available at https://github.com/yiheng-zhu/MKFGO.

MOTIVATION: Spatial clustering is a key analytical technique for exploring spatial transcriptomics data. Recent graph neural network-based methods have shown promise in spatial clustering but face notable challenges. One significant issue is that analyzing the functions and complex mechanisms of organisms from a single scale is difficult and most methods focus exclusively on the single-scale representation of transcriptomic data, potentially limiting the discriminative power of extracted features for spatial domain clustering. Furthermore, classical clustering algorithms are often applied directly to latent representation, making it a worthwhile endeavor to explore a tailored clustering method to further improve the accuracy of spatial domain annotation. RESULTS: To address these limitations, we propose m2ST, a novel dual multi-scale graph clustering method. m2ST first uses a multi-scale masked graph autoencoder to extract representations across different scales from spatial transcriptomic data. To effectively compress and distill meaningful knowledge embedded in the data, m2ST introduces a random masking mechanism for node features and uses a scaled cosine error as the loss function. Additionally, we introduce a tailored multi-scale clustering framework that integrates scale-common and scale-specific information exploration into the clustering process, achieving more robust annotation performance. Shannon entropy is finally utilized to dynamically adjust the importance of different scales. Extensive experiments on multiple spatial transcriptomic datasets demonstrate the superior performance of m2ST compared to existing methods. AVAILABILITY AND IMPLEMENTATION: https://github.com/BBKing49/m2ST.

Abstract The optic nerve carries signals from the eye to the brain, where they are interpreted as visual images. Optic neuropathy is a serious eye disease that can lead to the loss of vision in the affected eye. Identifying the risk factors for optic neuropathy from large patient data is crucial and challenging. Modern techniques can assist in recognizing these risk factors. For instance, data mining algorithms such as classification and association rules can discover knowledge from datasets in many real-world applications, particularly in the medical field. This study employed a decision tree algorithm known as J48 and an association rule algorithm called Apriori to analyze the collected data. The J48 algorithm achieved an accuracy of 90%, while the Apriori algorithm discovered 52 significant association rules with a confidence level above 80%. The goal of this study was to identify risk factors for optic neuropathy and explore the connection between optic nerve damage and other conditions. The proposed algorithms aim to reduce blindness rates and increase awareness of the risk factors associated with optic nerve damage by detecting hidden risk factors at an early stage. The study's findings show that some risk factors for optic neuropathy confirmed by medical trials are also detected by these algorithms, proving the effectiveness and applicability of data mining techniques in the medical field. Moreover, this study discovered new risk factors for optic nerve damage not previously found by medical trials. This knowledge will contribute to the early detection and prevention of blindness by recognizing risk factors for optic nerve damage.

The development of intelligent methods capable of predicting protein-ligand binding sites has become a popular research field. Recently, deep learning based methods have been proposed as a promising solution for this task. However, some limitations still exist. For example, the network structure is not optimized for predicting protein binding pockets, which limits the model's capabilities. To address the aforementioned challenges, a novel method called CATransUnetLPB is proposed, in which a new network structure named CATransUnet is designed. The proposed CATransUnet combines CNN and Transformer models to accurately segment binding pocket regions from protein 3D structures. It outperforms existing representative methods on three test sets, demonstrating the effectiveness of optimizing the deep network model for detecting protein ligand binding pockets. Furthermore, we conduct thorough analysis on applying data augmentation to protein data structure and confirm that such technique can enhance the model's generalization ability, thereby ensuring good performance on new protein structures. Moreover, experiments show that the predicted binding pockets from our model can complement the results obtained from other methods. This suggests that integrating our method with existing approaches could further improve the prediction of protein-ligand binding pockets.

Accurate prediction of protein subcellular localization is critical for understanding cellular functions and guiding drug design. However, current computational methods have limited and insufficient performance and as such, there exist few efficient vision learners based on self-supervised learning for extracting deep and informative features. To address it, we propose a novel bioimage-based method, termed PScL-SDNNMAE, to effectively predict the subcellular localizations of proteins in human cells. PScL-SDNNMAE first extracts classical features using traditional image descriptors. Next, the masked autoencoder (MAE) is first trained using the training image data and then used to extract the MAE-based deep features. In the feature selection phase, PScL-SDNNMAE applies the Analysis of Variance (ANOVA), Mutual Information (MI) and stepwise discriminant analysis (SDA) to select the optimal features from the classical feature sets. Finally, PScL-SDNNMAE trains the deep neural network (DNN) classifier using the super feature set generated by integrating all the classical optimal and MAE-based deep features. Extensive benchmark experiments including 10-fold cross-validation on the training dataset and independent test on the independent dataset illustrate more advanced performance and generalization capability of PScL-SDNNMAE than other existing state-of-the-art predictors. Moreover, the experiments also demonstrate the effectiveness of self-supervised learning methods in learning representations of IHC images, as well as the significant potential for pre-training on massive unlabeled datasets in the future.

Intelligent prediction of protein function is more efficient and less resource-consuming and has achieved significant progress in recent years. However, most of the current methods are performed solely based on the sequence information of proteins. These methods overlook information of other modalities that the proteins themselves possess, which makes it difficult to achieve the desired predicted results. Furthermore, a few existing methods based on multiple modal information fuse them in a simple splicing manner and fail to fully exploit the complementary relation between different modalities. To address the above-mentioned challenges, we propose Multi-modal Multi-attention fusion Features (DMMAFS), a method based on deep learning, to predict protein function. On the one hand, DMMAFS gains the semantic information embedded in the sequence itself through the self-attention learning of the sequence. On the other hand, DMMAFS employs the 3D structural information of proteins to compensate for the sequence information. Particularly, a S-C cross-modal cross-attention fusion network module is proposed that not only optimizes the weights of the semantic information but also efficiently fuses the sequence features with the structural information, thus avoiding the simple splicing of different modal features. Our experimental results demonstrate that the proposed DMMAFS outperforms the state-of-the-art methods in protein function prediction.

Single-cell transcriptomics is a powerful approach for characterizing gene transcription at cellular resolution. This approach requires efficient computational pipelines to undertake essential tasks, including clustering, dimensionality reduction, imputation, and denoising. Currently, most such pipelines undertake these computational tasks separately without considering the interdependence among these tasks. Here, we present an advanced pipeline, MUSIC-GCN, by employing a graph convolutional neural (GCN) network and autoencoder to perform multi-task single-cell RNA-sequencing (scRNA-seq) data analysis. The rationale is that multiple related tasks can be carried out simultaneously to enable enhanced learning and more effective representations through the 'sharing of knowledge' regarding individual tasks. Benchmarking experiments using various scRNA-seq datasets show that MUSIC-GCN can achieve a competitive performance on multi-tasks when benchmarked with state-of-the-art approaches.

Single-cell RNA sequencing (scRNA-seq) techniques are emerging to revolutionize modern biomedical sciences by providing a detailed landscape of individual cells. However, these methods often lack crucial spatial localization information. To address this gap, spatial transcriptomic technologies have developed, enabling gene expression profiling while mapping cells spatial information. Yet, the gene throughput in spatial transcriptomic technologies makes it challenging to characterize whole-transcriptome-level data for single cells in space. In this context, approaches for predicting the spatial distribution of genes are still under development. Here, we present GCNgene, a novel method to predict the spatial distribution of the undetected RNA transcripts, through integrating spatial and scRNA-seq datasets. GCNgene leverages a graph convolutional network to embed spatial transcriptomics data and then applies a learned rule to reconstruct gene expression by combining the reference single-cell data with the calculated cell-type proportions. Ultimately, this learned paradigm enables accurate predictions of gene expression levels.

Accurate identification of protein-nucleotide binding residues is essential for protein functional annotation and drug discovery. Advancements in computational methods for predicting binding residues from protein sequences have significantly improved predictive accuracy. However, it remains a challenge for current methodologies to extract discriminative features and assimilate heterogeneous data from different nucleotide binding residues. To address this, we introduce NucMoMTL, a novel predictor specifically designed for identifying protein-nucleotide binding residues. Specifically, NucMoMTL leverages a pre-trained language model for robust sequence embedding and utilizes deep multi-task and multi-scale learning within parameter-based orthogonal constraints to extract shared representations, capitalizing on auxiliary information from diverse nucleotides binding residues. Evaluation of NucMoMTL on the benchmark datasets demonstrates that it outperforms state-of-the-art methods, achieving an average AUROC and AUPRC of 0.961 and 0.566, respectively. NucMoMTL can be explored as a reliable computational tool for identifying protein-nucleotide binding residues and facilitating drug discovery.

The recent advances of single-cell RNA sequencing (scRNA-seq) have enabled reliable profiling of gene expression at the single-cell level, providing opportunities for accurate inference of gene regulatory networks (GRNs) on scRNA-seq data. Most methods for inferring GRNs suffer from the inability to eliminate transitive interactions or necessitate expensive computational resources. To address these, we present a novel method, termed GMFGRN, for accurate graph neural network (GNN)-based GRN inference from scRNA-seq data. GMFGRN employs GNN for matrix factorization and learns representative embeddings for genes. For transcription factor-gene pairs, it utilizes the learned embeddings to determine whether they interact with each other. The extensive suite of benchmarking experiments encompassing eight static scRNA-seq datasets alongside several state-of-the-art methods demonstrated mean improvements of 1.9 and 2.5% over the runner-up in area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC). In addition, across four time-series datasets, maximum enhancements of 2.4 and 1.3% in AUROC and AUPRC were observed in comparison to the runner-up. Moreover, GMFGRN requires significantly less training time and memory consumption, with time and memory consumed <10% compared to the second-best method. These findings underscore the substantial potential of GMFGRN in the inference of GRNs. It is publicly available at https://github.com/Lishuoyy/GMFGRN.

Accurate prediction of transcription factor binding sites (TFBSs) is essential for understanding gene regulation mechanisms and the etiology of diseases. Despite numerous advances in deep learning for predicting TFBSs, their performance can still be enhanced. In this study, we propose MLSNet, a novel deep learning architecture designed specifically to predict TFBSs. MLSNet innovatively integrates multisize convolutional fusion with long short-term memory (LSTM) networks to effectively capture DNA-sparse higher-order sequence features. Further, MLSNet incorporates super token attention and Bi-LSTM to systematically extract and integrate higher-order DNA shape features. Experimental results on 165 ChIP-seq (chromatin immunoprecipitation followed by sequencing) datasets indicate that MLSNet consistently outperforms several state-of-the-art algorithms in the prediction of TFBSs. Specifically, MLSNet reports average metrics: 0.8306 for ACC, 0.8992 for AUROC, and 0.9035 for AUPRC, surpassing the second-best methods by 1.82%, 1.68%, and 1.54%, respectively. This research delineates the effectiveness of combining multi-size convolutional layers with LSTM and DNA shape-based features in enhancing predictive accuracy. Moreover, this study comprehensively assesses the variability in model performance across different cell lines and transcription factors. The source code of MLSNet is available at https://github.com/minghaidea/MLSNet.

Efficient and accurate recognition of protein-DNA interactions is vital for understanding the molecular mechanisms of related biological processes and further guiding drug discovery. Although the current experimental protocols are the most precise way to determine protein-DNA binding sites, they tend to be labor-intensive and time-consuming. There is an immediate need to design efficient computational approaches for predicting DNA-binding sites. Here, we proposed ULDNA, a new deep-learning model, to deduce DNA-binding sites from protein sequences. This model leverages an LSTM-attention architecture, embedded with three unsupervised language models that are pre-trained on large-scale sequences from multiple database sources. To prove its effectiveness, ULDNA was tested on 229 protein chains with experimental annotation of DNA-binding sites. Results from computational experiments revealed that ULDNA significantly improves the accuracy of DNA-binding site prediction in comparison with 17 state-of-the-art methods. In-depth data analyses showed that the major strength of ULDNA stems from employing three transformer language models. Specifically, these language models capture complementary feature embeddings with evolution diversity, in which the complex DNA-binding patterns are buried. Meanwhile, the specially crafted LSTM-attention network effectively decodes evolution diversity-based embeddings as DNA-binding results at the residue level. Our findings demonstrated a new pipeline for predicting DNA-binding sites on a large scale with high accuracy from protein sequence alone.

MOTIVATION: Drug-target interaction (DTI) prediction refers to the prediction of whether a given drug molecule will bind to a specific target and thus exert a targeted therapeutic effect. Although intelligent computational approaches for drug target prediction have received much attention and made many advances, they are still a challenging task that requires further research. The main challenges are manifested as follows: (i) most graph neural network-based methods only consider the information of the first-order neighboring nodes (drug and target) in the graph, without learning deeper and richer structural features from the higher-order neighboring nodes. (ii) Existing methods do not consider both the sequence and structural features of drugs and targets, and each method is independent of each other, and cannot combine the advantages of sequence and structural features to improve the interactive learning effect. RESULTS: To address the above challenges, a Multi-view Integrated learning Network that integrates Deep learning and Graph Learning (MINDG) is proposed in this study, which consists of the following parts: (i) a mixed deep network is used to extract sequence features of drugs and targets, (ii) a higher-order graph attention convolutional network is proposed to better extract and capture structural features, and (iii) a multi-view adaptive integrated decision module is used to improve and complement the initial prediction results of the above two networks to enhance the prediction performance. We evaluate MINDG on two dataset and show it improved DTI prediction performance compared to state-of-the-art baselines. AVAILABILITY AND IMPLEMENTATION: https://github.com/jnuaipr/MINDG.

Recently, mask-fill-based 3D Molecular Generation (MG) methods have become very popular in virtual drug design. However, the existing MG methods ignore the chemical properties of atoms and contain inappropriate atomic position training data, which limits their generation capability. To mitigate the above issues, this paper presents a novel mask-fill-based 3D molecule generation model driven by atomic chemical properties (APMG). Specifically, we construct a new attention-MPNN-based encoder and introduce the electronic information into atom representations to enrich chemical properties. Also, a multi-functional classifier is designed to predict the electronic information of each generated atom, guiding the type prediction of elements and bonds. By design, the proposed method uses the chemical properties of atoms and their correlations for high-quality molecule generation. Second, to optimize the atomic position training data, we propose a novel atomic training position generation approach using the Chi-Square distribution. We evaluate our APMG method on the CrossDocked dataset and visualize the docking states of the pockets and generated molecules. The obtained results demonstrate the superiority and merits of APMG over the state-of-the-art approaches.

N4-acetylcytidine (ac4C) is a post-transcriptional modification in mRNA that is critical in mRNA translation in terms of stability and regulation. In the past few years, numerous approaches employing convolutional neural networks (CNN) and Transformer have been proposed for the identification of ac4C sites, with each variety of approaches processing distinct characteristics. CNN-based methods excel at extracting local features and positional information, whereas Transformer-based ones stands out in establishing long-range dependencies and generating global representations. Given the importance of both local and global features in mRNA ac4C sites identification, we propose a novel method termed TransC-ac4C which combines CNN and Transformer together for enhancing the feature extraction capability and improving the identification accuracy. Five different feature encoding strategies (One-hot, NCP, ND, EIIP, and K-mer) are employed to generate the mRNA sequence representations, in which way the sequence attributes and physical and chemical properties of the sequences can be embedded. To strengthen the relevance of features, we construct a novel feature fusion method. Firstly, the CNN is employed to process five single features, stitch them together and feed them to the Transformer layer. Then, our approach employs CNN to extract local features and Transformer subsequently to establish global long-range dependencies among extracted features. We use 5-fold cross-validation to evaluate the model, and the evaluation indicators are significantly improved. The prediction accuracy of the two datasets is as high as 81.42% and 80.69%, respectively. It demonstrates the stronger competitiveness and generalization performance of our model.

Accurate identification of antifreeze proteins (AFPs) is crucial in developing biomimetic synthetic anti-icing materials and low-temperature organ preservation materials. Although numerous machine learning-based methods have been proposed for AFPs prediction, the complex and diverse nature of AFPs limits the prediction performance of existing methods. In this study, we propose AFP-Deep, a new deep learning method to predict antifreeze proteins by integrating embedding from protein sequences with pre-trained protein language models and evolutionary contexts with hybrid feature extraction networks. The experimental results demonstrated that the main advantage of AFP-Deep is its utilization of pre-trained protein language models, which can extract discriminative global contextual features from protein sequences. Additionally, the hybrid deep neural networks designed for protein language models and evolutionary context feature extraction enhance the correlation between embeddings and antifreeze pattern. The performance evaluation results show that AFP-Deep achieves superior performance compared to state-of-the-art models on benchmark datasets, achieving an AUPRC of 0.724 and 0.924, respectively.

RNA N6-methyladenosine is a prevalent and abundant type of RNA modification that exerts significant influence on diverse biological processes. To date, numerous computational approaches have been developed for predicting methylation, with most of them ignoring the correlations of different encoding strategies and failing to explore the adaptability of various attention mechanisms for methylation identification. To solve the above issues, we proposed an innovative framework for predicting RNA m6A modification site, termed BLAM6A-Merge. Specifically, it utilized a multimodal feature fusion strategy to combine the classification results of four features and Blastn tool. Apart from this, different attention mechanisms were employed for extracting higher-level features on specific features after the screening process. Extensive experiments on 12 benchmarking datasets demonstrated that BLAM6A-Merge achieved superior performance (average AUC: 0.849 for the full transcript mode and 0.784 for the mature mRNA mode). Notably, the Blastn tool was employed for the first time in the identification of methylation sites.

Determining the pathogenicity and functional impact (i.e. gain-of-function; GOF or loss-of-function; LOF) of a variant is vital for unraveling the genetic level mechanisms of human diseases. To provide a 'one-stop' framework for the accurate identification of pathogenicity and functional impact of variants, we developed a two-stage deep-learning-based computational solution, termed VPatho, which was trained using a total of 9619 pathogenic GOF/LOF and 138 026 neutral variants curated from various databases. A total number of 138 variant-level, 262 protein-level and 103 genome-level features were extracted for constructing the models of VPatho. The development of VPatho consists of two stages: (i) a random under-sampling multi-scale residual neural network (ResNet) with a newly defined weighted-loss function (RUS-Wg-MSResNet) was proposed to predict variants' pathogenicity on the gnomAD_NV + GOF/LOF dataset; and (ii) an XGBOD model was constructed to predict the functional impact of the given variants. Benchmarking experiments demonstrated that RUS-Wg-MSResNet achieved the highest prediction performance with the weights calculated based on the ratios of neutral versus pathogenic variants. Independent tests showed that both RUS-Wg-MSResNet and XGBOD achieved outstanding performance. Moreover, assessed using variants from the CAGI6 competition, RUS-Wg-MSResNet achieved superior performance compared to state-of-the-art predictors. The fine-trained XGBOD models were further used to blind test the whole LOF data downloaded from gnomAD and accordingly, we identified 31 nonLOF variants that were previously labeled as LOF/uncertain variants. As an implementation of the developed approach, a webserver of VPatho is made publicly available at http://csbio.njust.edu.cn/bioinf/vpatho/ to facilitate community-wide efforts for profiling and prioritizing the query variants with respect to their pathogenicity and functional impact.

Single-cell RNA sequencing (scRNA-seq) has significantly accelerated the experimental characterization of distinct cell lineages and types in complex tissues and organisms. Cell-type annotation is of great importance in most of the scRNA-seq analysis pipelines. However, manual cell-type annotation heavily relies on the quality of scRNA-seq data and marker genes, and therefore can be laborious and time-consuming. Furthermore, the heterogeneity of scRNA-seq datasets poses another challenge for accurate cell-type annotation, such as the batch effect induced by different scRNA-seq protocols and samples. To overcome these limitations, here we propose a novel pipeline, termed TripletCell, for cross-species, cross-protocol and cross-sample cell-type annotation. We developed a cell embedding and dimension-reduction module for the feature extraction (FE) in TripletCell, namely TripletCell-FE, to leverage the deep metric learning-based algorithm for the relationships between the reference gene expression matrix and the query cells. Our experimental studies on 21 datasets (covering nine scRNA-seq protocols, two species and three tissues) demonstrate that TripletCell outperformed state-of-the-art approaches for cell-type annotation. More importantly, regardless of protocols or species, TripletCell can deliver outstanding and robust performance in annotating different types of cells. TripletCell is freely available at https://github.com/liuyan3056/TripletCell. We believe that TripletCell is a reliable computational tool for accurately annotating various cell types using scRNA-seq data and will be instrumental in assisting the generation of novel biological hypotheses in cell biology.

MOTIVATION: Over the past decades, a variety of in silico methods have been developed to predict protein subcellular localization within cells. However, a common and major challenge in the design and development of such methods is how to effectively utilize the heterogeneous feature sets extracted from bioimages. In this regards, limited efforts have been undertaken. RESULTS: We propose a new two-level stacked autoencoder network (termed 2L-SAE-SM) to improve its performance by integrating the heterogeneous feature sets. In particular, in the first level of 2L-SAE-SM, each optimal heterogeneous feature set is fed to train our designed stacked autoencoder network (SAE-SM). All the trained SAE-SMs in the first level can output the decision sets based on their respective optimal heterogeneous feature sets, known as 'intermediate decision' sets. Such intermediate decision sets are then ensembled using the mean ensemble method to generate the 'intermediate feature' set for the second-level SAE-SM. Using the proposed framework, we further develop a novel predictor, referred to as PScL-2LSAESM, to characterize image-based protein subcellular localization. Extensive benchmarking experiments on the latest benchmark training and independent test datasets collected from the human protein atlas databank demonstrate the effectiveness of the proposed 2L-SAE-SM framework for the integration of heterogeneous feature sets. Moreover, performance comparison of the proposed PScL-2LSAESM with current state-of-the-art methods further illustrates that PScL-2LSAESM clearly outperforms the existing state-of-the-art methods for the task of protein subcellular localization. AVAILABILITY AND IMPLEMENTATION: https://github.com/csbio-njust-edu/PScL-2LSAESM. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: CircRNAs play a critical regulatory role in physiological processes, and the abnormal expression of circRNAs can mediate the processes of diseases. Therefore, exploring circRNAs-disease associations is gradually becoming an important area of research. Due to the high cost of validating circRNA-disease associations using traditional wet-lab experiments, novel computational methods based on machine learning are gaining more and more attention in this field. However, current computational methods suffer to insufficient consideration of latent features in circRNA-disease interactions. RESULTS: In this study, a multilayer attention neural graph-based collaborative filtering (MLNGCF) is proposed. MLNGCF first enhances multiple biological information with autoencoder as the initial features of circRNAs and diseases. Then, by constructing a central network of different diseases and circRNAs, a multilayer cooperative attention-based message propagation is performed on the central network to obtain the high-order features of circRNAs and diseases. A neural network-based collaborative filtering is constructed to predict the unknown circRNA-disease associations and update the model parameters. Experiments on the benchmark datasets demonstrate that MLNGCF outperforms state-of-the-art methods, and the prediction results are supported by the literature in the case studies. AVAILABILITY AND IMPLEMENTATION: The source codes and benchmark datasets of MLNGCF are available at https://github.com/ABard0/MLNGCF.

MOTIVATION: RNA N6-methyladenosine (m6A) in Homo sapiens plays vital roles in a variety of biological functions. Precise identification of m6A modifications is thus essential to elucidation of their biological functions and underlying molecular-level mechanisms. Currently available high-throughput single-nucleotide-resolution m6A modification data considerably accelerated the identification of RNA modification sites through the development of data-driven computational methods. Nevertheless, existing methods have limitations in terms of the coverage of single-nucleotide-resolution cell lines and have poor capability in model interpretations, thereby having limited applicability. RESULTS: In this study, we present CLSM6A, comprising a set of deep learning-based models designed for predicting single-nucleotide-resolution m6A RNA modification sites across eight different cell lines and three tissues. Extensive benchmarking experiments are conducted on well-curated datasets and accordingly, CLSM6A achieves superior performance than current state-of-the-art methods. Furthermore, CLSM6A is capable of interpreting the prediction decision-making process by excavating critical motifs activated by filters and pinpointing highly concerned positions in both forward and backward propagations. CLSM6A exhibits better portability on similar cross-cell line/tissue datasets, reveals a strong association between highly activated motifs and high-impact motifs, and demonstrates complementary attributes of different interpretation strategies. AVAILABILITY AND IMPLEMENTATION: The webserver is available at http://csbio.njust.edu.cn/bioinf/clsm6a. The datasets and code are available at https://github.com/zhangying-njust/CLSM6A/.

Recently, deep learning has become the mainstream methodology for Compound-Protein Interaction (CPI) prediction. However, the existing compound-protein feature extraction methods have some issues that limit their performance. First, graph networks are widely used for structural compound feature extraction, but the chemical properties of a compound depend on functional groups rather than graphic structure. Besides, the existing methods lack capabilities in extracting rich and discriminative protein features. Last, the compound-protein features are usually simply combined for CPI prediction, without considering information redundancy and effective feature mining. To address the above issues, we propose a novel CPInformer method. Specifically, we extract heterogeneous compound features, including structural graph features and functional class fingerprints, to reduce prediction errors caused by similar structural compounds. Then, we combine local and global features using dense connections to obtain multi-scale protein features. Last, we apply ProbSparse self-attention to protein features, under the guidance of compound features, to eliminate information redundancy, and to improve the accuracy of CPInformer. More importantly, the proposed method identifies the activated local regions that link a CPI, providing a good visualisation for the CPI state. The results obtained on five benchmarks demonstrate the merits and superiority of CPInformer over the state-of-the-art approaches.

It has been demonstrated that RNA modifications play essential roles in multiple biological processes. Accurate identification of RNA modifications in the transcriptome is critical for providing insights into the biological functions and mechanisms. Many tools have been developed for predicting RNA modifications at single-base resolution, which employ conventional feature engineering methods that focus on feature design and feature selection processes that require extensive biological expertise and may introduce redundant information. With the rapid development of artificial intelligence technologies, end-to-end methods are favorably received by researchers. Nevertheless, each well-trained model is only suitable for a specific RNA methylation modification type for nearly all of these approaches. In this study, we present MRM-BERT by feeding task-specific sequences into the powerful BERT (Bidirectional Encoder Representations from Transformers) model and implementing fine-tuning, which exhibits competitive performance to the state-of-the-art methods. MRM-BERT avoids repeated de novo training of the model and can predict multiple RNA modifications such as pseudouridine, m6A, m5C, and m1A in Mus musculus, Arabidopsis thaliana, and Saccharomyces cerevisiae. In addition, we analyse the attention heads to provide high attention regions for the prediction, and conduct saturated in silico mutagenesis of the input sequences to discover potential changes of RNA modifications, which can better assist researchers in their follow-up research.

More than 6000 human diseases have been recorded to be caused by non-synonymous single nucleotide polymorphisms (nsSNPs). Rapid and accurate prediction of pathogenic nsSNPs can improve our understanding of the principle and design of new drugs, which remains an unresolved challenge. In the present work, a new computational approach, termed MSRes-MutP, is proposed based on ResNet blocks with multi-scale kernel size to predict disease-associated nsSNPs. By feeding the serial concatenation of the extracted four types of features, the performance of MSRes-MutP does not obviously improve. To address this, a second model FFMSRes-MutP is developed, which utilizes deep feature fusion strategy and multi-scale 2D-ResNet and 1D-ResNet blocks to extract relevant two-dimensional features and physicochemical properties. FFMSRes-MutP with the concatenated features achieves a better performance than that with individual features. The performance of FFMSRes-MutP is benchmarked on five different datasets. It achieves the Matthew's correlation coefficient (MCC) of 0.593 and 0.618 on the PredictSNP and MMP datasets, which are 0.101 and 0.210 higher than that of the existing best method PredictSNP1. When tested on the HumDiv and HumVar datasets, it achieves MCC of 0.9605 and 0.9507, and area under curve (AUC) of 0.9796 and 0.9748, which are 0.1747 and 0.2669, 0.0853 and 0.1335, respectively, higher than the existing best methods PolyPhen-2 and FATHMM (weighted). In addition, on blind test using a third-party dataset, FFMSRes-MutP performs as the second-best predictor (with MCC and AUC of 0.5215 and 0.7633, respectively), when compared with the other four predictors. Extensive benchmarking experiments demonstrate that FFMSRes-MutP achieves effective feature fusion and can be explored as a useful approach for predicting disease-associated nsSNPs. The webserver is freely available at http://csbio.njust.edu.cn/bioinf/ffmsresmutp/ for academic use.

Accurate identification of transcription factor binding sites is of great significance in understanding gene expression, biological development and drug design. Although a variety of methods based on deep-learning models and large-scale data have been developed to predict transcription factor binding sites in DNA sequences, there is room for further improvement in prediction performance. In addition, effective interpretation of deep-learning models is greatly desirable. Here we present MAResNet, a new deep-learning method, for predicting transcription factor binding sites on 690 ChIP-seq datasets. More specifically, MAResNet combines the bottom-up and top-down attention mechanisms and a state-of-the-art feed-forward network (ResNet), which is constructed by stacking attention modules that generate attention-aware features. In particular, the multi-scale attention mechanism is utilized at the first stage to extract rich and representative sequence features. We further discuss the attention-aware features learned from different attention modules in accordance with the changes as the layers go deeper. The features learned by MAResNet are also visualized through the TMAP tool to illustrate that the method can extract the unique characteristics of transcription factor binding sites. The performance of MAResNet is extensively tested on 690 test subsets with an average AUC of 0.927, which is higher than that of the current state-of-the-art methods. Overall, this study provides a new and useful framework for the prediction of transcription factor binding sites by combining the funnel attention modules with the residual network.

Circular RNA (circRNA) is closely involved in physiological and pathological processes of many diseases. Discovering the associations between circRNAs and diseases is of great significance. Due to the high-cost to verify the circRNA-disease associations by wet-lab experiments, computational approaches for predicting the associations become a promising research direction. In this paper, we propose a method, MDGF-MCEC, based on multi-view dual attention graph convolution network (GCN) with cooperative ensemble learning to predict circRNA-disease associations. First, MDGF-MCEC constructs two disease relation graphs and two circRNA relation graphs based on different similarities. Then, the relation graphs are fed into a multi-view GCN for representation learning. In order to learn high discriminative features, a dual-attention mechanism is introduced to adjust the contribution weights, at both channel level and spatial level, of different features. Based on the learned embedding features of diseases and circRNAs, nine different feature combinations between diseases and circRNAs are treated as new multi-view data. Finally, we construct a multi-view cooperative ensemble classifier to predict the associations between circRNAs and diseases. Experiments conducted on the CircR2Disease database demonstrate that the proposed MDGF-MCEC model achieves a high area under curve of 0.9744 and outperforms the state-of-the-art methods. Promising results are also obtained from experiments on the circ2Disease and circRNADisease databases. Furthermore, the predicted associated circRNAs for hepatocellular carcinoma and gastric cancer are supported by the literature. The code and dataset of this study are available at https://github.com/ABard0/MDGF-MCEC.

MOTIVATION: Accurately identifying protein-ATP binding poses is significantly valuable for both basic structure biology and drug discovery. Although many docking methods have been designed, most of them require a user-defined binding site and are difficult to achieve a high-quality protein-ATP docking result. It is critical to develop a protein-ATP-specific blind docking method without user-defined binding sites. RESULTS: Here, we present ATPdock, a template-based method for docking ATP into protein. For each query protein, if no pocket site is given, ATPdock first identifies its most potential pocket using ATPbind, an ATP-binding site predictor; then, the template pocket, which is most similar to the given or identified pocket, is searched from the database of pocket-ligand structures using APoc, a pocket structural alignment tool; thirdly, the rough docking pose of ATP (rdATP) is generated using LS-align, a ligand structural alignment tool, to align the initial ATP pose to the template ligand corresponding to template pocket; finally, the Metropolis Monte Carlo simulation is used to fine-tune the rdATP under the guidance of AutoDock Vina energy function. Benchmark tests show that ATPdock significantly outperforms other state-of-the-art methods in docking accuracy. AVAILABILITY AND IMPLEMENTATION: https://jun-csbio.github.io/atpdock/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Characterization of protein subcellular localization has become an important and long-standing task in bioinformatics and computational biology, which provides valuable information for elucidating various cellular functions of proteins and guiding drug design. RESULTS: Here, we develop a novel bioimage-based computational approach, termed PScL-DDCFPred, to accurately predict protein subcellular localizations in human tissues. PScL-DDCFPred first extracts multiview image features, including global and local features, as base or pure features; next, it applies a new integrative feature selection method based on stepwise discriminant analysis and generalized discriminant analysis to identify the optimal feature sets from the extracted pure features; Finally, a classifier based on deep neural network (DNN) and deep-cascade forest (DCF) is established. Stringent 10-fold cross-validation tests on the new protein subcellular localization training dataset, constructed from the human protein atlas databank, illustrates that PScL-DDCFPred achieves a better performance than several existing state-of-the-art methods. Moreover, the independent test set further illustrates the generalization capability and superiority of PScL-DDCFPred over existing predictors. In-depth analysis shows that the excellent performance of PScL-DDCFPred can be attributed to three critical factors, namely the effective combination of the DNN and DCF models, complementarity of global and local features, and use of the optimal feature sets selected by the integrative feature selection algorithm. AVAILABILITY AND IMPLEMENTATION: https://github.com/csbio-njust-edu/PScL-DDCFPred. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Accurate identification of protein function is critical to elucidate life mechanisms and design new drugs. We proposed a novel deep-learning method, ATGO, to predict Gene Ontology (GO) attributes of proteins through a triplet neural-network architecture embedded with pre-trained language models from protein sequences. The method was systematically tested on 1068 non-redundant benchmarking proteins and 3328 targets from the third Critical Assessment of Protein Function Annotation (CAFA) challenge. Experimental results showed that ATGO achieved a significant increase of the GO prediction accuracy compared to the state-of-the-art approaches in all aspects of molecular function, biological process, and cellular component. Detailed data analyses showed that the major advantage of ATGO lies in the utilization of pre-trained transformer language models which can extract discriminative functional pattern from the feature embeddings. Meanwhile, the proposed triplet network helps enhance the association of functional similarity with feature similarity in the sequence embedding space. In addition, it was found that the combination of the network scores with the complementary homology-based inferences could further improve the accuracy of the predicted models. These results demonstrated a new avenue for high-accuracy deep-learning function prediction that is applicable to large-scale protein function annotations from sequence alone.

Cell-penetrating peptides (CPPs) are special peptides capable of carrying a variety of bioactive molecules, such as genetic materials, short interfering RNAs and nanoparticles, into cells. Recently, research on CPP has gained substantial interest from researchers, and the biological mechanisms of CPPS have been assessed in the context of safe drug delivery agents and therapeutic applications. Correct identification and synthesis of CPPs using traditional biochemical methods is an extremely slow, expensive and laborious task particularly due to the large volume of unannotated peptide sequences accumulating in the World Bank repository. Hence, a powerful bioinformatics predictor that rapidly identifies CPPs with a high recognition rate is urgently needed. To date, numerous computational methods have been developed for CPP prediction. However, the available machine-learning (ML) tools are unable to distinguish both the CPPs and their uptake efficiencies. This study aimed to develop a two-layer deep learning framework named DeepCPPred to identify both CPPs in the first phase and peptide uptake efficiency in the second phase. The DeepCPPred predictor first uses four types of descriptors that cover evolutionary, energy estimation, reduced sequence and amino-acid contact information. Then, the extracted features are optimized through the elastic net algorithm and fed into a cascade deep forest algorithm to build the final CPP model. The proposed method achieved 99.45 percent overall accuracy with the CPP924 benchmark dataset in the first layer and 95.43 percent accuracy in the second layer with the CPPSite3 dataset using a 5-fold cross-validation test. Thus, our proposed bioinformatics tool surpassed all the existing state-of-the-art sequence-based CPP approaches.

Protein-DNA interactions play an important role in diverse biological processes. Accurately identifying protein-DNA binding residues is a critical but challenging task for protein function annotations and drug design. Although wet-lab experimental methods are the most accurate way to identify protein-DNA binding residues, they are time consuming and labor intensive. There is an urgent need to develop computational methods to rapidly and accurately predict protein-DNA binding residues. In this study, we propose a novel sequence-based method, named PredDBR, for predicting DNA-binding residues. In PredDBR, for each query protein, its position-specific frequency matrix (PSFM), predicted secondary structure (PSS), and predicted probabilities of ligand-binding residues (PPLBR) are first generated as three feature sources. Secondly, for each feature source, the sliding window technique is employed to extract the matrix-format feature of each residue. Then, we design two strategies, i.e., square root (SR) and average (AVE), to separately transform PSFM-based and two predicted feature source-based, i.e., PSS-based and PPLBR-based, matrix-format features of each residue into three corresponding cube-format features. Finally, after serially combining the three cube-format features, the ensemble classifier is generated via applying bagging strategy to multiple base classifiers built by the framework of 2D convolutional neural network. The computational experimental results demonstrate that the proposed PredDBR achieves an average overall accuracy of 93.7% and a Mathew's correlation coefficient of 0.405 on two independent validation datasets and outperforms several state-of-the-art sequenced-based protein-DNA binding residue predictors. The PredDBR web-server is available at https://jun-csbio.github.io/PredDBR/.

Accurate short-term traffic flow prediction plays an important role in the field of modern Intelligent Transportation Systems. Since various uncontrollable factors (e.g.weather, traffic jams or accidents), collected traffic data inevitably contain outliers. This makes it challenge to achieve satisfactory results for traffic flow prediction. Least Squares Twin Support Vector Regression (LSTSVR) has been shown to provide a powerful potential in nonlinear prediction problems. This is especially true when using appropriate heuristic algorithms to determine the parameters of nonlinear LSTSVR. In view of this, a novel LSTSVR model based on the robust$\text{L}_{2,\mathrm {p}}$-norm ($0< p\le 2$) distance is proposed to alleviate the negative effect of traffic data with outliers, called PLSTSVR. An iterative algorithm is designed to solve the optimization problem of PLSTSVR, which has great potential for solving other relevant optimization problems. To search the parameters of constructed PLSTSVR, this paper constructs two traffic flow prediction models based on PLSTSVR and heuristic algorithms (Fruit Fly Optimization Algorithm and Particle Swarm Optimization), called PLSTSVR-FOA and PLSTSVR-PSO. Extensive experiments demonstrate that the constructed models are more effective and robust than other competing models in various experimental settings.

DNA N6-methyladenine is an important type of DNA modification that plays important roles in multiple biological processes. Despite the recent progress in developing DNA 6mA site prediction methods, several challenges remain to be addressed. For example, although the hand-crafted features are interpretable, they contain redundant information that may bias the model training and have a negative impact on the trained model. Furthermore, although deep learning (DL)-based models can perform feature extraction and classification automatically, they lack the interpretability of the crucial features learned by those models. As such, considerable research efforts have been focused on achieving the trade-off between the interpretability and straightforwardness of DL neural networks. In this study, we develop two new DL-based models for improving the prediction of N6-methyladenine sites, termed LA6mA and AL6mA, which use bidirectional long short-term memory to respectively capture the long-range information and self-attention mechanism to extract the key position information from DNA sequences. The performance of the two proposed methods is benchmarked and evaluated on the two model organisms Arabidopsis thaliana and Drosophila melanogaster. On the two benchmark datasets, LA6mA achieves an area under the receiver operating characteristic curve (AUROC) value of 0.962 and 0.966, whereas AL6mA achieves an AUROC value of 0.945 and 0.941, respectively. Moreover, an in-depth analysis of the attention matrix is conducted to interpret the important information, which is hidden in the sequence and relevant for 6mA site prediction. The two novel pipelines developed for DNA 6mA site prediction in this work will facilitate a better understanding of the underlying principle of DL-based DNA methylation site prediction and its future applications.

Protein fold recognition is a critical step toward protein structure and function prediction, aiming at providing the most likely fold type of the query protein. In recent years, the development of deep learning (DL) technique has led to massive advances in this important field, and accordingly, the sensitivity of protein fold recognition has been dramatically improved. Most DL-based methods take an intermediate bottleneck layer as the feature representation of proteins with new fold types. However, this strategy is indirect, inefficient and conditional on the hypothesis that the bottleneck layer's representation is assumed as a good representation of proteins with new fold types. To address the above problem, in this work, we develop a new computational framework by combining triplet network and ensemble DL. We first train a DL-based model, termed FoldNet, which employs triplet loss to train the deep convolutional network. FoldNet directly optimizes the protein fold embedding itself, making the proteins with the same fold types be closer to each other than those with different fold types in the new protein embedding space. Subsequently, using the trained FoldNet, we implement a new residue-residue contact-assisted predictor, termed FoldTR, which improves protein fold recognition. Furthermore, we propose a new ensemble DL method, termed FSD_XGBoost, which combines protein fold embedding with the other two discriminative fold-specific features extracted by two DL-based methods SSAfold and DeepFR. The Top 1 sensitivity of FSD_XGBoost increases to 74.8% at the fold level, which is ~9% higher than that of the state-of-the-art method. Together, the results suggest that fold-specific features extracted by different DL methods complement with each other, and their combination can further improve fold recognition at the fold level. The implemented web server of FoldTR and benchmark datasets are publicly available at http://csbio.njust.edu.cn/bioinf/foldtr/.

As an essential task in protein structure and function prediction, protein fold recognition has attracted increasing attention. The majority of the existing machine learning-based protein fold recognition approaches strongly rely on handcrafted features, which depict the characteristics of different protein folds; however, effective feature extraction methods still represent the bottleneck for further performance improvement of protein fold recognition. As a powerful feature extractor, deep convolutional neural network (DCNN) can automatically extract discriminative features for fold recognition without human intervention, which has demonstrated an impressive performance on protein fold recognition. Despite the encouraging progress, DCNN often acts as a black box, and as such, it is challenging for users to understand what really happens in DCNN and why it works well for protein fold recognition. In this study, we explore the intrinsic mechanism of DCNN and explain why it works for protein fold recognition using a visual explanation technique. More specifically, we first trained a VGGNet-based DCNN model, termed VGGNet-FE, which can extract fold-specific features from the predicted protein residue-residue contact map for protein fold recognition. Subsequently, based on the trained VGGNet-FE, we implemented a new contact-assisted predictor, termed VGGfold, for protein fold recognition; we then visualized what features were extracted by each of the convolutional layers in VGGNet-FE using a deconvolution technique. Furthermore, we visualized the high-level semantic information, termed fold-discriminative region, of a predicted contact map from the localization map obtained from the last convolutional layer of VGGNet-FE. It is visually confirmed that VGGNet-FE could effectively extract distinct fold-discriminative regions for different types of protein folds, thereby accounting for the improved performance of VGGfold for protein fold recognition. In summary, this study is of great significance for both understanding the working principle of DCNNs in protein fold recognition and exploring the relationship between the predicted protein contact map and protein tertiary structure. This proposed visualization method is flexible and applicable to address other DCNN-based bioinformatics and computational biology questions. The online web server of VGGfold is freely available at http://csbio.njust.edu.cn/bioinf/vggfold/.

Knowledge of the specificity of DNA-protein binding is crucial for understanding the mechanisms of gene expression, regulation and gene therapy. In recent years, deep-learning-based methods for predicting DNA-protein binding from sequence data have achieved significant success. Nevertheless, the current state-of-the-art computational methods have some drawbacks associated with the use of limited datasets with insufficient experimental data. To address this, we propose a novel transfer learning-based method, termed SAResNet, which combines the self-attention mechanism and residual network structure. More specifically, the attention-driven module captures the position information of the sequence, while the residual network structure guarantees that the high-level features of the binding site can be extracted. Meanwhile, the pre-training strategy used by SAResNet improves the learning ability of the network and accelerates the convergence speed of the network during transfer learning. The performance of SAResNet is extensively tested on 690 datasets from the ChIP-seq experiments with an average AUC of 92.0%, which is 4.4% higher than that of the best state-of-the-art method currently available. When tested on smaller datasets, the predictive performance is more clearly improved. Overall, we demonstrate that the superior performance of DNA-protein binding prediction on DNA sequences can be achieved by combining the attention mechanism and residual structure, and a novel pipeline is accordingly developed. The proposed methodology is generally applicable and can be used to address any other sequence classification problems.

Protein subcellular localization plays a crucial role in characterizing the function of proteins and understanding various cellular processes. Therefore, accurate identification of protein subcellular location is an important yet challenging task. Numerous computational methods have been proposed to predict the subcellular location of proteins. However, most existing methods have limited capability in terms of the overall accuracy, time consumption and generalization power. To address these problems, in this study, we developed a novel computational approach based on human protein atlas (HPA) data, referred to as PScL-HDeep, for accurate and efficient image-based prediction of protein subcellular location in human tissues. We extracted different handcrafted and deep learned (by employing pretrained deep learning model) features from different viewpoints of the image. The step-wise discriminant analysis (SDA) algorithm was applied to generate the optimal feature set from each original raw feature set. To further obtain a more informative feature subset, support vector machine-based recursive feature elimination with correlation bias reduction (SVM-RFE + CBR) feature selection algorithm was applied to the integrated feature set. Finally, the classification models, namely support vector machine with radial basis function (SVM-RBF) and support vector machine with linear kernel (SVM-LNR), were learned on the final selected feature set. To evaluate the performance of the proposed method, a new gold standard benchmark training dataset was constructed from the HPA databank. PScL-HDeep achieved the maximum performance on 10-fold cross validation test on this dataset and showed a better efficacy over existing predictors. Furthermore, we also illustrated the generalization ability of the proposed method by conducting a stringent independent validation test.

Antimicrobial peptides (AMPs) are a unique and diverse group of molecules that play a crucial role in a myriad of biological processes and cellular functions. AMP-related studies have become increasingly popular in recent years due to antimicrobial resistance, which is becoming an emerging global concern. Systematic experimental identification of AMPs faces many difficulties due to the limitations of current methods. Given its significance, more than 30 computational methods have been developed for accurate prediction of AMPs. These approaches show high diversity in their data set size, data quality, core algorithms, feature extraction, feature selection techniques and evaluation strategies. Here, we provide a comprehensive survey on a variety of current approaches for AMP identification and point at the differences between these methods. In addition, we evaluate the predictive performance of the surveyed tools based on an independent test data set containing 1536 AMPs and 1536 non-AMPs. Furthermore, we construct six validation data sets based on six different common AMP databases and compare different computational methods based on these data sets. The results indicate that amPEPpy achieves the best predictive performance and outperforms the other compared methods. As the predictive performances are affected by the different data sets used by different methods, we additionally perform the 5-fold cross-validation test to benchmark different traditional machine learning methods on the same data set. These cross-validation results indicate that random forest, support vector machine and eXtreme Gradient Boosting achieve comparatively better performances than other machine learning methods and are often the algorithms of choice of multiple AMP prediction tools.

X-ray crystallography is the major approach for determining atomic-level protein structures. Because not all proteins can be easily crystallized, accurate prediction of protein crystallization propensity provides critical help in guiding experimental design and improving the success rate of X-ray crystallography experiments. This study has developed a new machine-learning-based pipeline that uses a newly developed deep-cascade forest (DCF) model with multiple types of sequence-based features to predict protein crystallization propensity. Based on the developed pipeline, two new protein crystallization propensity predictors, denoted as DCFCrystal and MDCFCrystal, have been implemented. DCFCrystal is a multistage predictor that can estimate the success propensities of the three individual steps (production of protein material, purification and production of crystals) in the protein crystallization process. MDCFCrystal is a single-stage predictor that aims to estimate the probability that a protein will pass through the entire crystallization process. Moreover, DCFCrystal is designed for general proteins, whereas MDCFCrystal is specially designed for membrane proteins, which are notoriously difficult to crystalize. DCFCrystal and MDCFCrystal were separately tested on two benchmark datasets consisting of 12 289 and 950 proteins, respectively, with known crystallization results from various experimental records. The experimental results demonstrated that DCFCrystal and MDCFCrystal increased the value of Matthew's correlation coefficient by 199.7% and 77.8%, respectively, compared to the best of other state-of-the-art protein crystallization propensity predictors. Detailed analyses show that the major advantages of DCFCrystal and MDCFCrystal lie in the efficiency of the DCF model and the sensitivity of the sequence-based features used, especially the newly designed pseudo-predicted hybrid solvent accessibility (PsePHSA) feature, which improves crystallization recognition by incorporating sequence-order information with solvent accessibility of residues. Meanwhile, the new crystal-dataset constructions help to train the models with more comprehensive crystallization knowledge.

The topology of protein folds can be specified by the inter-residue contact-maps and accurate contact-map prediction can help ab initio structure folding. We developed TripletRes to deduce protein contact-maps from discretized distance profiles by end-to-end training of deep residual neural-networks. Compared to previous approaches, the major advantage of TripletRes is in its ability to learn and directly fuse a triplet of coevolutionary matrices extracted from the whole-genome and metagenome databases and therefore minimize the information loss during the course of contact model training. TripletRes was tested on a large set of 245 non-homologous proteins from CASP 11&12 and CAMEO experiments and outperformed other top methods from CASP12 by at least 58.4% for the CASP 11&12 targets and 44.4% for the CAMEO targets in the top-L long-range contact precision. On the 31 FM targets from the latest CASP13 challenge, TripletRes achieved the highest precision (71.6%) for the top-L/5 long-range contact predictions. It was also shown that a simple re-training of the TripletRes model with more proteins can lead to further improvement with precisions comparable to state-of-the-art methods developed after CASP13. These results demonstrate a novel efficient approach to extend the power of deep convolutional networks for high-accuracy medium- and long-range protein contact-map predictions starting from primary sequences, which are critical for constructing 3D structure of proteins that lack homologous templates in the PDB library.

This article presents a new Self-growing and Pruning Generative Adversarial Network (SP-GAN) for realistic image generation. In contrast to traditional GAN models, our SP-GAN is able to dynamically adjust the size and architecture of a network in the training stage by using the proposed self-growing and pruning mechanisms. To be more specific, we first train two seed networks as the generator and discriminator; each contains a small number of convolution kernels. Such small-scale networks are much easier and faster to train than large-capacity networks. Second, in the self-growing step, we replicate the convolution kernels of each seed network to augment the scale of the network, followed by fine-tuning the augmented/expanded network. More importantly, to prevent the excessive growth of each seed network in the self-growing stage, we propose a pruning strategy that reduces the redundancy of an augmented network, yielding the optimal scale of the network. Finally, we design a new adaptive loss function that is treated as a variable loss computational process for the training of the proposed SP-GAN model. By design, the hyperparameters of the loss function can dynamically adapt to different training stages. Experimental results obtained on a set of data sets demonstrate the merits of the proposed method, especially in terms of the stability and efficiency of network training. The source code of the proposed SP-GAN method is publicly available at https://github.com/Lambert-chen/SPGAN.git.

Accurately identifying DNA-binding proteins (DBPs) from protein sequence information is an important but challenging task for protein function annotations. In this paper, we establish a novel computational method, named TargetDBP, for accurately targeting DBPs from primary sequences. In TargetDBP, four single-view features, i.e., AAC (Amino Acid Composition), PsePSSM (Pseudo Position-Specific Scoring Matrix), PsePRSA (Pseudo Predicted Relative Solvent Accessibility), and PsePPDBS (Pseudo Predicted Probabilities of DNA-Binding Sites), are first extracted to represent different base features, respectively. Second, differential evolution algorithm is employed to learn the weights of four base features. Using the learned weights, we weightedly combine these base features to form the original super feature. An excellent subset of the super feature is then selected by using a suitable feature selection algorithm SVM-REF+CBR (Support Vector Machine Recursive Feature Elimination with Correlation Bias Reduction). Finally, the prediction model is learned via using support vector machine on the selected feature subset. We also construct a new gold-standard and non-redundant benchmark dataset from PDB database to evaluate and compare the proposed TargetDBP with other existing predictors. On this new dataset, TargetDBP can achieve higher performance than other state-of-the-art predictors. The TargetDBP web server and datasets are freely available at http://csbio.njust.edu.cn/bioinf/targetdbp/ for academic use.

Active learning algorithms attempt to overcome the labeling bottleneck by asking queries from large collection of unlabeled examples. Existing batch mode active learning algorithms sufferfrom three limitations: (1) The methods that are based on similarityfunction or optimizing certain diversity measurement, in which may lead to suboptimal performance and produce the selected set with redundant examples. (2) The models with assumption on data are hard in finding images that are both informative and representative. (3) The problem of noise labels has been an obstacle for algorithms. In this paper, we propose a novel active learning method that makes embeddings of labeled examples to those of unlabeled ones and back via deep neural networks. The active scheme makes correct association cycles that end up at the same class from that the association was started, which considers both the informativeness and representativeness of examples, as well as being robust to the noise labels. We apply our active learning method to semi-supervised classification and clustering. The submodular function is designed to reduce the redundancy of the selected examples. Specifically, we incorporate our batch mode active scheme into the classification approaches, in which the generalization ability is improved. For semi-supervised clustering, we try to use our active scheme for constraints to make fast convergence and perform better than unsupervised clustering. Finally, we apply our active learning method to data filtering. To validate the effectiveness of the proposed algorithms, extensive experiments are conducted on diversity benchmark datasets for different tasks, i.e., classification, clustering, and data filtering, and the experimental results demonstrate consistent and substantial improvements over the state-of-the-art approaches.

MOTIVATION: Contact-map of a protein sequence dictates the global topology of structural fold. Accurate prediction of the contact-map is thus essential to protein 3D structure prediction, which is particularly useful for the protein sequences that do not have close homology templates in the Protein Data Bank. RESULTS: We developed a new method, ResPRE, to predict residue-level protein contacts using inverse covariance matrix (or precision matrix) of multiple sequence alignments (MSAs) through deep residual convolutional neural network training. The approach was tested on a set of 158 non-homologous proteins collected from the CASP experiments and achieved an average accuracy of 50.6% in the top-L long-range contact prediction with L being the sequence length, which is 11.7% higher than the best of other state-of-the-art approaches ranging from coevolution coupling analysis to deep neural network training. Detailed data analyses show that the major advantage of ResPRE lies at the utilization of precision matrix that helps rule out transitional noises of contact-maps compared with the previously used covariance matrix. Meanwhile, the residual network with parallel shortcut layer connections increases the learning ability of deep neural network training. It was also found that appropriate collection of MSAs can further improve the accuracy of final contact-map predictions. The standalone package and online server of ResPRE are made freely available, which should bring important impact on protein structure and function modeling studies in particular for the distant- and non-homology protein targets. AVAILABILITY AND IMPLEMENTATION: https://zhanglab.ccmb.med.umich.edu/ResPRE and https://github.com/leeyang/ResPRE. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Motivation: Sequence-order independent structural comparison, also called structural alignment, of small ligand molecules is often needed for computer-aided virtual drug screening. Although many ligand structure alignment programs are proposed, most of them build the alignments based on rigid-body shape comparison which cannot provide atom-specific alignment information nor allow structural variation; both abilities are critical to efficient high-throughput virtual screening. Results: We propose a novel ligand comparison algorithm, LS-align, to generate fast and accurate atom-level structural alignments of ligand molecules, through an iterative heuristic search of the target function that combines inter-atom distance with mass and chemical bond comparisons. LS-align contains two modules of Rigid-LS-align and Flexi-LS-align, designed for rigid-body and flexible alignments, respectively, where a ligand-size independent, statistics-based scoring function is developed to evaluate the similarity of ligand molecules relative to random ligand pairs. Large-scale benchmark tests are performed on prioritizing chemical ligands of 102 protein targets involving 1 415 871 candidate compounds from the DUD-E (Database of Useful Decoys: Enhanced) database, where LS-align achieves an average enrichment factor (EF) of 22.0 at the 1% cutoff and the AUC score of 0.75, which are significantly higher than other state-of-the-art methods. Detailed data analyses show that the advanced performance is mainly attributed to the design of the target function that combines structural and chemical information to enhance the sensitivity of recognizing subtle difference of ligand molecules and the introduces of structural flexibility that help capture the conformational changes induced by the ligand-receptor binding interactions. These data demonstrate a new avenue to improve the virtual screening efficiency through the development of sensitive ligand structural alignments. Availability and implementation: http://zhanglab.ccmb.med.umich.edu/LS-align/. Supplementary information: Supplementary data are available at Bioinformatics online.

Accurate identification of the cancer types is essential to cancer diagnoses and treatments. Since cancer tissue and normal tissue have different gene expression, gene expression data can be used as an efficient feature source for cancer classification. However, accurate cancer classification directly using original gene expression profiles remains challenging due to the intrinsic high-dimension feature and the small size of the data samples. We proposed a new self-training subspace clustering algorithm under low-rank representation, called SSC-LRR, for cancer classification on gene expression data. Low-rank representation (LRR) is first applied to extract discriminative features from the high-dimensional gene expression data; the self-training subspace clustering (SSC) method is then used to generate the cancer classification predictions. The SSC-LRR was tested on two separate benchmark datasets in control with four state-of-the-art classification methods. It generated cancer classification predictions with an overall accuracy 89.7 percent and a general correlation 0.920, which are 18.9 and 24.4 percent higher than that of the best control method respectively. In addition, several genes (RNF114, HLA-DRB5, USP9Y, and PTPN20) were identified by SSC-LRR as new cancer identifiers that deserve further clinical investigation. Overall, the study demonstrated a new sensitive avenue to recognize cancer classifications from large-scale gene expression data.

Protein-DNA interactions are ubiquitous in a wide variety of biological processes. Correctly locating DNA-binding residues solely from protein sequences is an important but challenging task for protein function annotations and drug discovery, especially in the post-genomic era where large volumes of protein sequences have quickly accumulated. In this study, we report a new predictor, named TargetDNA, for targeting protein-DNA binding residues from primary sequences. TargetDNA uses a protein's evolutionary information and its predicted solvent accessibility as two base features and employs a centered linear kernel alignment algorithm to learn the weights for weightedly combining the two features. Based on the weightedly combined feature, multiple initial predictors with SVM as classifiers are trained by applying a random under-sampling technique to the original dataset, the purpose of which is to cope with the severe imbalance phenomenon that exists between the number of DNA-binding and non-binding residues. The final ensembled predictor is obtained by boosting the multiple initially trained predictors. Experimental simulation results demonstrate that the proposed TargetDNA achieves a high prediction performance and outperforms many existing sequence-based protein-DNA binding residue predictors. The TargetDNA web server and datasets are freely available at http://csbio.njust.edu.cn/bioinf/TargetDNA/ for academic use.

Disulfide connectivity is an important protein structural characteristic. Accurately predicting disulfide connectivity solely from protein sequence helps to improve the intrinsic understanding of protein structure and function, especially in the post-genome era where large volume of sequenced proteins without being functional annotated is quickly accumulated. In this study, a new feature extracted from the predicted protein 3D structural information is proposed and integrated with traditional features to form discriminative features. Based on the extracted features, a random forest regression model is performed to predict protein disulfide connectivity. We compare the proposed method with popular existing predictors by performing both cross-validation and independent validation tests on benchmark datasets. The experimental results demonstrate the superiority of the proposed method over existing predictors. We believe the superiority of the proposed method benefits from both the good discriminative capability of the newly developed features and the powerful modelling capability of the random forest. The web server implementation, called TargetDisulfide, and the benchmark datasets are freely available at: http://csbio.njust.edu.cn/bioinf/TargetDisulfide for academic use.

BACKGROUND: Vitamins are typical ligands that play critical roles in various metabolic processes. The accurate identification of the vitamin-binding residues solely based on a protein sequence is of significant importance for the functional annotation of proteins, especially in the post-genomic era, when large volumes of protein sequences are accumulating quickly without being functionally annotated. RESULTS: In this paper, a new predictor called TargetVita is designed and implemented for predicting protein-vitamin binding residues using protein sequences. In TargetVita, features derived from the position-specific scoring matrix (PSSM), predicted protein secondary structure, and vitamin binding propensity are combined to form the original feature space; then, several feature subspaces are selected by performing different feature selection methods. Finally, based on the selected feature subspaces, heterogeneous SVMs are trained and then ensembled for performing prediction. CONCLUSIONS: The experimental results obtained with four separate vitamin-binding benchmark datasets demonstrate that the proposed TargetVita is superior to the state-of-the-art vitamin-specific predictor, and an average improvement of 10% in terms of the Matthews correlation coefficient (MCC) was achieved over independent validation tests. The TargetVita web server and the datasets used are freely available for academic use at http://csbio.njust.edu.cn/bioinf/TargetVita or http://www.csbio.sjtu.edu.cn/bioinf/TargetVita.

Accurately identifying the protein-ligand binding sites or pockets is of significant importance for both protein function analysis and drug design. Although much progress has been made, challenges remain, especially when the 3D structures of target proteins are not available or no homology templates can be found in the library, where the template-based methods are hard to be applied. In this paper, we report a new ligand-specific template-free predictor called TargetS for targeting protein-ligand binding sites from primary sequences. TargetS first predicts the binding residues along the sequence with ligand-specific strategy and then further identifies the binding sites from the predicted binding residues through a recursive spatial clustering algorithm. Protein evolutionary information, predicted protein secondary structure, and ligand-specific binding propensities of residues are combined to construct discriminative features; an improved AdaBoost classifier ensemble scheme based on random undersampling is proposed to deal with the serious imbalance problem between positive (binding) and negative (nonbinding) samples. Experimental results demonstrate that TargetS achieves high performances and outperforms many existing predictors. TargetS web server and data sets are freely available at: http://www.csbio.sjtu.edu.cn/bioinf/TargetS/ for academic use.

BACKGROUND: Adenosine-5'-triphosphate (ATP) is one of multifunctional nucleotides and plays an important role in cell biology as a coenzyme interacting with proteins. Revealing the binding sites between protein and ATP is significantly important to understand the functionality of the proteins and the mechanisms of protein-ATP complex. RESULTS: In this paper, we propose a novel framework for predicting the proteins' functional residues, through which they can bind with ATP molecules. The new prediction protocol is achieved by combination of sequence evolutional information and bi-profile sampling of multi-view sequential features and the sequence derived structural features. The hypothesis for this strategy is single-view feature can only represent partial target's knowledge and multiple sources of descriptors can be complementary. CONCLUSIONS: Prediction performances evaluated by both 5-fold and leave-one-out jackknife cross-validation tests on two benchmark datasets consisting of 168 and 227 non-homologous ATP binding proteins respectively demonstrate the efficacy of the proposed protocol. Our experimental results also reveal that the residue structural characteristics of real protein-ATP binding sites are significant different from those normal ones, for example the binding residues do not show high solvent accessibility propensities, and the bindings prefer to occur at the conjoint points between different secondary structure segments. Furthermore, results also show that performance is affected by the imbalanced training datasets by testing multiple ratios between positive and negative samples in the experiments. Increasing the dataset scale is also demonstrated useful for improving the prediction performances.

The purpose of this paper is to present a new rough set model for generating negative rules from the incomplete information system. A negative rule indicates that if an object does not satisfy the attribute-value pairs in the condition part, then we can exclude the decision part from such object. The proposed rough set model is constructed on the basis of a difference relation. Such difference relation is a binary relation without any constraints. Moreover, to simplify the negative rules generated from the difference relation-based rough approximations, the concepts of lower, upper approximate and rough reducts are also proposed. Some numerical examples are employed to substantiate the conceptual arguments.