A Face Detection and Standardized Mask-Wearing Recognition Algorithm.

In the era of coronavirus disease (COVID-19), wearing a mask could effectively protect people from the risk of infection and largely reduce transmission in public places. To prevent the spread of the virus, instruments are needed in public places to monitor whether people are wearing masks, which has higher requirements for the accuracy and speed of detection algorithms. To meet the demand for high accuracy and real-time monitoring, we propose a single-stage approach based on YOLOv4 to identify the face and whether to regulate the wearing of masks. In this approach, we propose a new feature pyramidal network based on the attention mechanism to reduce the loss of object information that can be caused by sampling and pooling in convolutional neural networks. The network is able to deeply mine the feature map for spatial and communication factors, and the multi-scale feature fusion makes the feature map equipped with location and semantic information. Based on the complete intersection over union (CIoU), a penalty function based on the norm is proposed to improve positioning accuracy, which is more accurate at the detection of small objects; the new bounding box regression function is called Norm CIoU (NCIoU). This function is applicable to various object-detection bounding box regression tasks. A combination of the two functions to calculate the confidence loss is used to mitigate the problem of the algorithm bias towards determinating no objects in the image. Moreover, we provide a dataset for recognizing faces and masks (RFM) that includes 12,133 realistic images. The dataset contains three categories: face, standardized mask and non-standardized mask. Experiments conducted on the dataset demonstrate that the proposed approach achieves mAP@.5:.95 69.70% and AP75 73.80%, outperforming the compared methods.

Using Machine Learning Algorithms to Determine the Post-COVID State of a Person by Their Rhythmogram.

This study introduces a novel method for detecting the post-COVID state using ECG data. By leveraging a convolutional neural network, we identify "cardiospikes" present in the ECG data of individuals who have experienced a COVID-19 infection. With a test sample, we achieve an 87 percent accuracy in detecting these cardiospikes. Importantly, our research demonstrates that these observed cardiospikes are not artifacts of hardware-software signal distortions, but rather possess an inherent nature, indicating their potential as markers for COVID-specific modes of heart rhythm regulation. Additionally, we conduct blood parameter measurements on recovered COVID-19 patients and construct corresponding profiles. These findings contribute to the field of remote screening using mobile devices and heart rate telemetry for diagnosing and monitoring COVID-19.

A Review Paper about Deep Learning for Medical Image Analysis.

Medical imaging refers to the process of obtaining images of internal organs for therapeutic purposes such as discovering or studying diseases. The primary objective of medical image analysis is to improve the efficacy of clinical research and treatment options. Deep learning has revamped medical image analysis, yielding excellent results in image processing tasks such as registration, segmentation, feature extraction, and classification. The prime motivations for this are the availability of computational resources and the resurgence of deep convolutional neural networks. Deep learning techniques are good at observing hidden patterns in images and supporting clinicians in achieving diagnostic perfection. It has proven to be the most effective method for organ segmentation, cancer detection, disease categorization, and computer-assisted diagnosis. Many deep learning approaches have been published to analyze medical images for various diagnostic purposes. In this paper, we review the work exploiting current state-of-the-art deep learning approaches in medical image processing. We begin the survey by providing a synopsis of research works in medical imaging based on convolutional neural networks. Second, we discuss popular pretrained models and general adversarial networks that aid in improving convolutional networks' performance. Finally, to ease direct evaluation, we compile the performance metrics of deep learning models focusing on COVID-19 detection and child bone age prediction.

COVID-19 Vaccine Hesitancy: A Global Public Health and Risk Modelling Framework Using an Environmental Deep Neural Network, Sentiment Classification with Text Mining and Emotional Reactions from COVID-19 Vaccination Tweets.

Popular social media platforms, such as Twitter, have become an excellent source of information with their swift information dissemination. Individuals with different backgrounds convey their opinions through social media platforms. Consequently, these platforms have become a profound instrument for collecting enormous datasets. We believe that compiling, organizing, exploring, and analyzing data from social media platforms, such as Twitter, can offer various perspectives to public health organizations and decision makers in identifying factors that contribute to vaccine hesitancy. In this study, public tweets were downloaded daily from Tweeter using the Tweeter API. Before performing computation, the tweets were preprocessed and labeled. Vocabulary normalization was based on stemming and lemmatization. The NRCLexicon technique was deployed to convert the tweets into ten classes: positive sentiment, negative sentiment, and eight basic emotions (joy, trust, fear, surprise, anticipation, anger, disgust, and sadness). t-test was used to check the statistical significance of the relationships among the basic emotions. Our analysis shows that the p-values of joy-sadness, trust-disgust, fear-anger, surprise-anticipation, and negative-positive relations are close to zero. Finally, neural network architectures, including 1DCNN, LSTM, Multiple-Layer Perceptron, and BERT, were trained and tested in a COVID-19 multi-classification of sentiments and emotions (positive, negative, joy, sadness, trust, disgust, fear, anger, surprise, and anticipation). Our experiment attained an accuracy of 88.6% for 1DCNN at 1744 s, 89.93% accuracy for LSTM at 27,597 s, while MLP achieved an accuracy of 84.78% at 203 s. The study results show that the BERT model performed the best, with an accuracy of 96.71% at 8429 s.

COVID-19 Detection from Cough Recordings Using Bag-of-Words Classifiers.

Reliable detection of COVID-19 from cough recordings is evaluated using bag-of-words classifiers. The effect of using four distinct feature extraction procedures and four different encoding strategies is evaluated in terms of the Area Under Curve (AUC), accuracy, sensitivity, and F1-score. Additional studies include assessing the effect of both input and output fusion approaches and a comparative analysis against 2D solutions using Convolutional Neural Networks. Extensive experiments conducted on the COUGHVID and COVID-19 Sounds datasets indicate that sparse encoding yields the best performances, showing robustness against various combinations of feature type, encoding strategy, and codebook dimension parameters.

NHGNN-DTA: a node-adaptive hybrid graph neural network for interpretable drug-target binding affinity prediction.

MOTIVATION: Large-scale prediction of drug-target affinity (DTA) plays an important role in drug discovery. In recent years, machine learning algorithms have made great progress in DTA prediction by utilizing sequence or structural information of both drugs and proteins. However, sequence-based algorithms ignore the structural information of molecules and proteins, while graph-based algorithms are insufficient in feature extraction and information interaction. RESULTS: In this article, we propose NHGNN-DTA, a node-adaptive hybrid neural network for interpretable DTA prediction. It can adaptively acquire feature representations of drugs and proteins and allow information to interact at the graph level, effectively combining the advantages of both sequence-based and graph-based approaches. Experimental results have shown that NHGNN-DTA achieved new state-of-the-art performance. It achieved the mean squared error (MSE) of 0.196 on the Davis dataset (below 0.2 for the first time) and 0.124 on the KIBA dataset (3% improvement). Meanwhile, in the case of cold start scenario, NHGNN-DTA proved to be more robust and more effective with unseen inputs than baseline methods. Furthermore, the multi-head self-attention mechanism endows the model with interpretability, providing new exploratory insights for drug discovery. The case study on Omicron variants of SARS-CoV-2 illustrates the efficient utilization of drug repurposing in COVID-19. AVAILABILITY AND IMPLEMENTATION: The source code and data are available at https://github.com/hehh77/NHGNN-DTA.

A tree based eXtreme Gradient Boosting (XGBoost) machine learning model to forecast the annual rice production in Bangladesh.

In this study, we attempt to anticipate annual rice production in Bangladesh (1961-2020) using both the Autoregressive Integrated Moving Average (ARIMA) and the eXtreme Gradient Boosting (XGBoost) methods and compare their respective performances. On the basis of the lowest Corrected Akaike Information Criteria (AICc) values, a significant ARIMA (0, 1, 1) model with drift was chosen based on the findings. The drift parameter value shows that the production of rice positively trends upward. Thus, the ARIMA (0, 1, 1) model with drift was found to be significant. On the other hand, the XGBoost model for time series data was developed by changing the tunning parameters frequently with the greatest result. The four prominent error measures, such as mean absolute error (MAE), mean percentage error (MPE), root mean square error (RMSE), and mean absolute percentage error (MAPE), were used to assess the predictive performance of each model. We found that the error measures of the XGBoost model in the test set were comparatively lower than those of the ARIMA model. Comparatively, the MAPE value of the test set of the XGBoost model (5.38%) was lower than that of the ARIMA model (7.23%), indicating that XGBoost performs better than ARIMA at predicting the annual rice production in Bangladesh. Hence, the XGBoost model performs better than the ARIMA model in predicting the annual rice production in Bangladesh. Therefore, based on the better performance, the study forecasted the annual rice production for the next 10 years using the XGBoost model. According to our predictions, the annual rice production in Bangladesh will vary from 57,850,318 tons in 2021 to 82,256,944 tons in 2030. The forecast indicated that the amount of rice produced annually in Bangladesh will increase in the years to come.

Artificial neural network based prediction of the lung tissue involvement as an independent in-hospital mortality and mechanical ventilation risk factor in COVID-19.

INTRODUCTION: During COVID-19 pandemic, artificial neural network (ANN) systems have been providing aid for clinical decisions. However, to achieve optimal results, these models should link multiple clinical data points to simple models. This study aimed to model the in-hospital mortality and mechanical ventilation risk using a two step approach combining clinical variables and ANN-analyzed lung inflammation data. METHODS: A data set of 4317 COVID-19 hospitalized patients, including 266 patients requiring mechanical ventilation, was analyzed. Demographic and clinical data (including the length of hospital stay and mortality) and chest computed tomography (CT) data were collected. Lung involvement was analyzed using a trained ANN. The combined data were then analyzed using unadjusted and multivariate Cox proportional hazards models. RESULTS: Overall in-hospital mortality associated with ANN-assigned percentage of the lung involvement (hazard ratio [HR]: 5.72, 95% confidence interval [CI]: 4.4-7.43, p < 0.001 for the patients with >50% of lung tissue affected by COVID-19 pneumonia), age category (HR: 5.34, 95% CI: 3.32-8.59 for cases >80 years, p < 0.001), procalcitonin (HR: 2.1, 95% CI: 1.59-2.76, p < 0.001, C-reactive protein level (CRP) (HR: 2.11, 95% CI: 1.25-3.56, p = 0.004), glomerular filtration rate (eGFR) (HR: 1.82, 95% CI: 1.37-2.42, p < 0.001) and troponin (HR: 2.14, 95% CI: 1.69-2.72, p < 0.001). Furthermore, the risk of mechanical ventilation is also associated with ANN-based percentage of lung inflammation (HR: 13.2, 95% CI: 8.65-20.4, p < 0.001 for patients with >50% involvement), age, procalcitonin (HR: 1.91, 95% CI: 1.14-3.2, p = 0.14, eGFR (HR: 1.82, 95% CI: 1.2-2.74, p = 0.004) and clinical variables, including diabetes (HR: 2.5, 95% CI: 1.91-3.27, p < 0.001), cardiovascular and cerebrovascular disease (HR: 3.16, 95% CI: 2.38-4.2, p < 0.001) and chronic pulmonary disease (HR: 2.31, 95% CI: 1.44-3.7, p < 0.001). CONCLUSIONS: ANN-based lung tissue involvement is the strongest predictor of unfavorable outcomes in COVID-19 and represents a valuable support tool for clinical decisions.

Generating and Weighting Semantically Consistent Sample Pairs for Ultrasound Contrastive Learning.

Well-annotated medical datasets enable deep neural networks (DNNs) to gain strong power in extracting lesion-related features. Building such large and well-designed medical datasets is costly due to the need for high-level expertise. Model pre-training based on ImageNet is a common practice to gain better generalization when the data amount is limited. However, it suffers from the domain gap between natural and medical images. In this work, we pre-train DNNs on ultrasound (US) domains instead of ImageNet to reduce the domain gap in medical US applications. To learn US image representations based on unlabeled US videos, we propose a novel meta-learning-based contrastive learning method, namely Meta Ultrasound Contrastive Learning (Meta-USCL). To tackle the key challenge of obtaining semantically consistent sample pairs for contrastive learning, we present a positive pair generation module along with an automatic sample weighting module based on meta-learning. Experimental results on multiple computer-aided diagnosis (CAD) problems, including pneumonia detection, breast cancer classification, and breast tumor segmentation, show that the proposed self-supervised method reaches state-of-the-art (SOTA). The codes are available at https://github.com/Schuture/Meta-USCL.

AI-Driven De Novo Design and Molecular Modeling for Discovery of Small-Molecule Compounds as Potential Drug Candidates Targeting SARS-CoV-2 Main Protease.

Over the past three years, significant progress has been made in the development of novel promising drug candidates against COVID-19. However, SARS-CoV-2 mutations resulting in the emergence of new viral strains that can be resistant to the drugs used currently in the clinic necessitate the development of novel potent and broad therapeutic agents targeting different vulnerable spots of the viral proteins. In this study, two deep learning generative models were developed and used in combination with molecular modeling tools for de novo design of small molecule compounds that can inhibit the catalytic activity of SARS-CoV-2 main protease (Mpro), an enzyme critically important for mediating viral replication and transcription. As a result, the seven best scoring compounds that exhibited low values of binding free energy comparable with those calculated for two potent inhibitors of Mpro, via the same computational protocol, were selected as the most probable inhibitors of the enzyme catalytic site. In light of the data obtained, the identified compounds are assumed to present promising scaffolds for the development of new potent and broad-spectrum drugs inhibiting SARS-CoV-2 Mpro, an attractive therapeutic target for anti-COVID-19 agents.

Classification of nucleic acid amplification on ISFET arrays using spectrogram-based neural networks.

The COVID-19 pandemic has highlighted a significant research gap in the field of molecular diagnostics. This has brought forth the need for AI-based edge solutions that can provide quick diagnostic results whilst maintaining data privacy, security and high standards of sensitivity and specificity. This paper presents a novel proof-of-concept method to detect nucleic acid amplification using ISFET sensors and deep learning. This enables the detection of DNA and RNA on a low-cost and portable lab-on-chip platform for identifying infectious diseases and cancer biomarkers. We show that by using spectrograms to transform the signal to the time-frequency domain, image processing techniques can be applied to achieve the reliable classification of the detected chemical signals. Transformation to spectrograms is beneficial as it makes the data compatible with 2D convolutional neural networks and helps gain significant performance improvement over neural networks trained on the time domain data. The trained network achieves an accuracy of 84% with a size of 30kB making it suitable for deployment on edge devices. This facilitates a new wave of intelligent lab-on-chip platforms that combine microfluidics, CMOS-based chemical sensing arrays and AI-based edge solutions for more intelligent and rapid molecular diagnostics.

A genetic programming-based convolutional deep learning algorithm for identifying COVID-19 cases via X-ray images.

Evolutionary algorithms have been successfully employed to find the best structure for many learning algorithms including neural networks. Due to their flexibility and promising results, Convolutional Neural Networks (CNNs) have found their application in many image processing applications. The structure of CNNs greatly affects the performance of these algorithms both in terms of accuracy and computational cost, thus, finding the best architecture for these networks is a crucial task before they are employed. In this paper, we develop a genetic programming approach for the optimization of CNN structure in diagnosing COVID-19 cases via X-ray images. A graph representation for CNN architecture is proposed and evolutionary operators including crossover and mutation are specifically designed for the proposed representation. The proposed architecture of CNNs is defined by two sets of parameters, one is the skeleton which determines the arrangement of the convolutional and pooling operators and their connections and one is the numerical parameters of the operators which determine the properties of these operators like filter size and kernel size. The proposed algorithm in this paper optimizes the skeleton and the numerical parameters of the CNN architectures in a co-evolutionary scheme. The proposed algorithm is used to identify covid-19 cases via X-ray images.

Evaluation of stacked ensemble model performance to predict clinical outcomes: A COVID-19 study.

BACKGROUND: The application of machine learning (ML) to analyze clinical data with the goal to predict patient outcomes has garnered increasing attention. Ensemble learning has been used in conjunction with ML to improve predictive performance. Although stacked generalization (stacking), a type of heterogeneous ensemble of ML models, has emerged in clinical data analysis, it remains unclear how to define the best model combinations for strong predictive performance. This study develops a methodology to evaluate the performance of "base" learner models and their optimized combination using "meta" learner models in stacked ensembles to accurately assess performance in the context of clinical outcomes. METHODS: De-identified COVID-19 data was obtained from the University of Louisville Hospital, where a retrospective chart review was performed from March 2020 to November 2021. Three differently-sized subsets using features from the overall dataset were chosen to train and evaluate ensemble classification performance. The number of base learners chosen from several algorithm families coupled with a complementary meta learner was varied from a minimum of 2 to a maximum of 8. Predictive performance of these combinations was evaluated in terms of mortality and severe cardiac event outcomes using area-under-the-receiver-operating-characteristic (AUROC), F1, balanced accuracy, and kappa. RESULTS: The results highlight the potential to accurately predict clinical outcomes, such as severe cardiac events with COVID-19, from routinely acquired in-hospital patient data. Meta learners Generalized Linear Model (GLM), Multi-Layer Perceptron (MLP), and Partial Least Squares (PLS) had the highest AUROC for both outcomes, while K-Nearest Neighbors (KNN) had the lowest. Performance trended lower in the training set as the number of features increased, and exhibited less variance in both training and validation across all feature subsets as the number of base learners increased. CONCLUSION: This study offers a methodology to robustly evaluate ensemble ML performance when analyzing clinical data.

Coagulation parameters predict COVID-19-related thrombosis in a neural network with a positive predictive value of 98.

Thrombosis is a major clinical complication of COVID-19 infection. COVID-19 patients show changes in coagulation factors that indicate an important role for the coagulation system in the pathogenesis of COVID-19. However, the multifactorial nature of thrombosis complicates the prediction of thrombotic events based on a single hemostatic variable. We developed and validated a neural net for the prediction of COVID-19-related thrombosis. The neural net was developed based on the hemostatic and general (laboratory) variables of 149 confirmed COVID-19 patients from two cohorts: at the time of hospital admission (cohort 1 including 133 patients) and at ICU admission (cohort 2 including 16 patients). Twenty-six patients suffered from thrombosis during their hospital stay: 19 patients in cohort 1 and 7 patients in cohort 2. The neural net predicts COVID-19 related thrombosis based on C-reactive protein (relative importance 14%), sex (10%), thrombin generation (TG) time-to-tail (10%), α2-Macroglobulin (9%), TG curve width (9%), thrombin-α2-Macroglobulin complexes (9%), plasmin generation lag time (8%), serum IgM (8%), TG lag time (7%), TG time-to-peak (7%), thrombin-antithrombin complexes (5%), and age (5%). This neural net can predict COVID-19-thrombosis at the time of hospital admission with a positive predictive value of 98%-100%.

Multimodal representation learning for predicting molecule-disease relations.

MOTIVATION: Predicting molecule-disease indications and side effects is important for drug development and pharmacovigilance. Comprehensively mining molecule-molecule, molecule-disease and disease-disease semantic dependencies can potentially improve prediction performance. METHODS: We introduce a Multi-Modal REpresentation Mapping Approach to Predicting molecular-disease relations (M2REMAP) by incorporating clinical semantics learned from electronic health records (EHR) of 12.6 million patients. Specifically, M2REMAP first learns a multimodal molecule representation that synthesizes chemical property and clinical semantic information by mapping molecule chemicals via a deep neural network onto the clinical semantic embedding space shared by drugs, diseases and other common clinical concepts. To infer molecule-disease relations, M2REMAP combines multimodal molecule representation and disease semantic embedding to jointly infer indications and side effects. RESULTS: We extensively evaluate M2REMAP on molecule indications, side effects and interactions. Results show that incorporating EHR embeddings improves performance significantly, for example, attaining an improvement over the baseline models by 23.6% in PRC-AUC on indications and 23.9% on side effects. Further, M2REMAP overcomes the limitation of existing methods and effectively predicts drugs for novel diseases and emerging pathogens. AVAILABILITY AND IMPLEMENTATION: The code is available at https://github.com/celehs/M2REMAP, and prediction results are provided at https://shiny.parse-health.org/drugs-diseases-dev/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Static Hand Gesture Recognition Using Capacitive Sensing and Machine Learning.

Automated hand gesture recognition is a key enabler of Human-to-Machine Interfaces (HMIs) and smart living. This paper reports the development and testing of a static hand gesture recognition system using capacitive sensing. Our system consists of a 6×18 array of capacitive sensors that captured five gestures-Palm, Fist, Middle, OK, and Index-of five participants to create a dataset of gesture images. The dataset was used to train Decision Tree, Naïve Bayes, Multi-Layer Perceptron (MLP) neural network, and Convolutional Neural Network (CNN) classifiers. Each classifier was trained five times; each time, the classifier was trained using four different participants' gestures and tested with one different participant's gestures. The MLP classifier performed the best, achieving an average accuracy of 96.87% and an average F1 score of 92.16%. This demonstrates that the proposed system can accurately recognize hand gestures and that capacitive sensing is a viable method for implementing a non-contact, static hand gesture recognition system.

Assessing lake water quality during COVID-19 era using geospatial techniques and artificial neural network model.

The present study evaluates the impact of the COVID-19 lockdown on the water quality of a tropical lake (East Kolkata Wetland or EKW, India) along with seasonal change using Landsat 8 and 9 images of the Google Earth Engine (GEE) cloud computing platform. The research focuses on detecting, monitoring, and predicting water quality in the EKW region using eight parameters-normalized suspended material index (NSMI), suspended particular matter (SPM), total phosphorus (TP), electrical conductivity (EC), chlorophyll-α, floating algae index (FAI), turbidity, Secchi disk depth (SDD), and two water quality indices such as Carlson tropic state index (CTSI) and entropy­weighted water quality index (EWQI). The results demonstrate that SPM, turbidity, EC, TP, and SDD improved while the FAI and chlorophyll-α increased during the lockdown period due to the stagnation of water as well as a reduction in industrial and anthropogenic pollution. Moreover, the prediction of EWQI using an artificial neural network indicates that the overall water quality will improve more if the lockdown period is sustained for another 3 years. The outcomes of the study will help the stakeholders develop effective regulations and strategies for the timely restoration of lake water quality.

Predicting locations of cryptic pockets from single protein structures using the PocketMiner graph neural network.

Cryptic pockets expand the scope of drug discovery by enabling targeting of proteins currently considered undruggable because they lack pockets in their ground state structures. However, identifying cryptic pockets is labor-intensive and slow. The ability to accurately and rapidly predict if and where cryptic pockets are likely to form from a structure would greatly accelerate the search for druggable pockets. Here, we present PocketMiner, a graph neural network trained to predict where pockets are likely to open in molecular dynamics simulations. Applying PocketMiner to single structures from a newly curated dataset of 39 experimentally confirmed cryptic pockets demonstrates that it accurately identifies cryptic pockets (ROC-AUC: 0.87) >1,000-fold faster than existing methods. We apply PocketMiner across the human proteome and show that predicted pockets open in simulations, suggesting that over half of proteins thought to lack pockets based on available structures likely contain cryptic pockets, vastly expanding the potentially druggable proteome.

An accessible infrastructure for artificial intelligence using a Docker-based JupyterLab in Galaxy.

BACKGROUND: Artificial intelligence (AI) programs that train on large datasets require powerful compute infrastructure consisting of several CPU cores and GPUs. JupyterLab provides an excellent framework for developing AI programs, but it needs to be hosted on such an infrastructure to enable faster training of AI programs using parallel computing. FINDINGS: An open-source, docker-based, and GPU-enabled JupyterLab infrastructure is developed that runs on the public compute infrastructure of Galaxy Europe consisting of thousands of CPU cores, many GPUs, and several petabytes of storage to rapidly prototype and develop end-to-end AI projects. Using a JupyterLab notebook, long-running AI model training programs can also be executed remotely to create trained models, represented in open neural network exchange (ONNX) format, and other output datasets in Galaxy. Other features include Git integration for version control, the option of creating and executing pipelines of notebooks, and multiple dashboards and packages for monitoring compute resources and visualization, respectively. CONCLUSIONS: These features make JupyterLab in Galaxy Europe highly suitable for creating and managing AI projects. A recent scientific publication that predicts infected regions in COVID-19 computed tomography scan images is reproduced using various features of JupyterLab on Galaxy Europe. In addition, ColabFold, a faster implementation of AlphaFold2, is accessed in JupyterLab to predict the 3-dimensional structure of protein sequences. JupyterLab is accessible in 2 ways-one as an interactive Galaxy tool and the other by running the underlying Docker container. In both ways, long-running training can be executed on Galaxy's compute infrastructure. Scripts to create the Docker container are available under MIT license at https://github.com/usegalaxy-eu/gpu-jupyterlab-docker.

LigBind: Identifying Binding Residues for Over 1000 Ligands with Relation-Aware Graph Neural Networks.

Identifying the interactions between proteins and ligands is significant for drug discovery and design. Considering the diverse binding patterns of ligands, the ligand-specific methods are trained per ligand to predict binding residues. However, most of the existing ligand-specific methods ignore shared binding preferences among various ligands and generally only cover a limited number of ligands with a sufficient number of known binding proteins. In this study, we propose a relation-aware framework LigBind with graph-level pre-training to enhance the ligand-specific binding residue predictions for 1159 ligands, which can effectively cover the ligands with a few known binding proteins. LigBind first pre-trains a graph neural network-based feature extractor for ligand-residue pairs and relation-aware classifiers for similar ligands. Then, LigBind is fine-tuned with ligand-specific binding data, where a domain adaptive neural network is designed to automatically leverage the diversity and similarity of various ligand-binding patterns for accurate binding residue prediction. We construct ligand-specific benchmark datasets of 1159 ligands and 16 unseen ligands, which are used to evaluate the effectiveness of LigBind. The results demonstrate the LigBind's efficacy on large-scale ligand-specific benchmark datasets, and it generalizes well to unseen ligands. LigBind also enables accurate identification of the ligand-binding residues in the main protease, papain-like protease and the RNA-dependent RNA polymerase of SARS-CoV-2. The web server and source codes of LigBind are available at http://www.csbio.sjtu.edu.cn/bioinf/LigBind/ and https://github.com/YYingXia/LigBind/ for academic use.