A Survey on Applications of Artificial Intelligence in Fighting Against COVID-19
Jianguo Chen, Kenli Li, Zhaolei Zhang, Keqin Li, Philip S. Yu
AA Survey on Applications of Artificial Intelligence in FightingAgainst COVID-19
Jianguo Chen , Kenli Li , Zhaolei Zhang , Keqin Li , Philip S. Yu College of Computer Science and Electronic Engineering, Hunan University, Changsha, Hunan,410082, China. Donnelly Centre for Cellular and Biomolecular Research and Department of Computer Science,University of Toronto, Toronto, ON, 710049, Canada. Department of Computer Science, State University of New York, New Paltz, NY, 12561, USA. Department of Computer Science, University of Illinois at Chicago, Chicago, IL, 60607, USA.Correspinding authors: Kenli Li ([email protected]) and Keqin Li ([email protected]).
Abstract
The COVID-19 pandemic caused by the SARS-CoV-2 virus has spread rapidly worldwide, leading to aglobal outbreak. Most governments, enterprises, and scientific research institutions are participating inthe COVID-19 struggle to curb the spread of the pandemic. As a powerful tool against COVID-19,artificial intelligence (AI) technologies are widely used in combating this pandemic. In this survey, weinvestigate the main scope and contributions of AI in combating COVID-19 from the aspects of diseasedetection and diagnosis, virology and pathogenesis, drug and vaccine development, and epidemic andtransmission prediction. In addition, we summarize the available data and resources that can be used forAI-based COVID-19 research. Finally, the main challenges and potential directions of AI in fightingagainst COVID-19 are discussed. Currently, AI mainly focuses on medical image inspection, genomics,drug development, and transmission prediction, and thus AI still has great potential in this field. Thissurvey presents medical and AI researchers with a comprehensive view of the existing and potentialapplications of AI technology in combating COVID-19 with the goal of inspiring researches to continueto maximize the advantages of AI and big data to fight COVID-19.
Severe Acute Respiratory Syndrome Corona-Virus 2 (SARS-CoV-2) is an emerging human infectiouscoronavirus. Since December 2019, coronavirus disease 2019 (COVID-19) caused by SARS-CoV-2 wasfirst reported in China and later in most of the countries in the world [10, 59, 220]. The World HealthOrganization (WHO) announced on January 30, 2020, that the outbreak was a Public Health Emergencyof International Concern (PHEIC), and confirmed COVID-19 as a pandemic on March 11, 2020. As ofMay 16, 2020, this disease has been reported in 216 countries or regions around the world and hasresulted in serious consequences including 4,525,497 confirmed COVID-19 cases and 307,395 deaths [216].Most governments, enterprises, and scientific research institutions are fighting COVID-19 from allaspects to curb the spread of the disease [49, 55, 58, 144]. Various stakeholders from different institutionsand backgrounds have provided abundant resources and capabilities to support this diseasebattle [16, 34, 53, 61, 131, 139, 201]. The current scope of combating COVID-19 includes virology andpathogenesis, disease diagnosis, epidemiology, therapeutics, and social control [95, 236]. The virology,origin and classification, physicochemical properties, receptor interactions, cell entry, genomic variation,and ecology in terms of SARS-CoV-2 are thoroughly studied [7, 77, 123, 123, 217, 242]. Nucleic acidtesting, serologic diagnosis, and medical imaging (chest X-ray or CT imaging) are the main disease1/36 a r X i v : . [ q - b i o . Q M ] J u l etection and diagnosis methods at present [30, 83, 106, 198, 227]. In terms of pathogenesis, topics such asvirus entry and spread, pathological findings, and immune response are thefocus [146, 196, 206, 207, 211, 224]. In terms of epidemiology, extensive research has been conducted on thesource and spectrum of infection, clinical features, epidemiological characteristics, epidemic prediction,and transmission route tracking [27, 115, 221, 222]. Potential therapeutics of COVID-19 include intensivecare, drug development, and vaccine development [36, 118, 150, 200]. In addition, communicationprediction and social isolation are the current main social control methods [1, 99, 157, 203, 235].Artificial intelligence (AI) is defined as a technology that allows computers to imitate humanintelligence to process things, including Machine Learning (ML), knowledge graphs, natural languageprocessing, human-computer interaction, computer vision, biometrics, virtual reality, and augmentedreality [26, 114, 189]. As a subset of AI, ML is a type of algorithms that can automatically analyze andobtain rules from data and use these rules for prediction or decision [156, 228, 229]. ML can besubdivided into traditional ML and deep learning (DL). Traditional ML methods include logisticregression, decision tree, random forest, K-nearest neighbor, Adaboost, K-means clustering, densityclustering, hidden Markov models, support vector machine, Naive Bayes, etc [22, 125, 142]. DL is a subsetof ML, and is a learning method for building deep structural models. Typical DL algorithms includedeep belief networks, convolutional neural networks, restricted Boltzmann machines, and recurrentneural networks [66, 186]. In addition, ML techniques also include transfer learning, active learning, andevolutionary learning. In recent years, AI technologies have achieved technical breakthroughs and arewidely used in various fields of intelligent medicine, including medical image inspection, disease-assisteddiagnosis, surgery, hospital management, and medical big data integration [189]. In addition, AI isactively explored in emerging fields such as surgical robots, wearable devices, new drug discovery,precision medicine, epidemics prevention and control, and gene sequencing.Encouragingly, in the short period of time since the outbreak of COVID-19, research in the industry,medical, and scientific fields has successfully used advanced AI technologies in the COVID-19 battle andhas achieved significant progress. For example, AI supports COVID-19 diagnosis through medical imageinspection and provides noninvasive detection solutions to prevent medical personnel from contractinginfections [4, 26, 114]. AI is used in virology research to analyze the structure of SARS-CoV-2-relatedproteins and predict new compounds that can be used for drug and vaccine development [23, 136, 243]. Inaddition, AI has achieved virus source tracking through genomics research and successfully discoveredthe relationships between SARS-CoV-2 and the bat virus, as well as SARS-CoV and the Middle Eastrespiratory syndrome-related coronavirus (MERS-CoV) [46, 145, 164]. Moreover, AI learns large-scaleCOVID-19 case data and social media data to construct epidemic transmission models to accuratelypredict the time of disease outbreak, transmission route, transmission range, andimpact [116, 147, 152, 171]. AI is also widely used in epidemic prevention and social control, such asairport security inspections, patient trajectory tracking, and epidemic visualization [163, 163].In this survey, we present a comprehensive view of the landscape and contributions of AI incombating COVID-19. The main scope of AI in COVID-19 research includes the aspects of diseasedetection and diagnosis, virology and pathogenesis, drug and vaccine development, and epidemic andtransmission prediction. Note that due to the rapid development of the COVID-19 epidemic, we citedmany preprinted references for a comprehensive investigation, where these references still need to beassessed on their accuracy and quality through peer review. The main scope of AI in combatingCOVID-19 is summarized in Fig. 1.The rest of the paper is structured as follows. Sections 2 to 5 discuss the four main scope of AI againstCOVID-19. Section 6 summarizes the available data and resources to support COVID-19 research.Section 7 highlights the challenges and potential directions in this field. Finally, Section 8 presents theconclusion. Table 1 gives the abbreviations and descriptions of the AI methods used in this survey. Diagnosis of virus infection is an important part of COVID-19 research. The current detection anddiagnosis methods for SARS-CoV-2 virus and COVID-19 disease mainly include nucleic acid testing,2/36 igure 1.
Main scope of AI in fighting against COVID-19. We collected 1273 online publications relatedto COVID-19, SARS-CoV-2, and 2019-nCoV from databases such as Nature, Elsevier, Google Scholar,arxiv, biorxiv, and medRxiv. Then, we filter out 267 papers that explicitly use AI methods.
Table 1.
Abbreviations and descriptions of the AI methods used in this survey.
Abbreviation Description Abbreviation DescriptionCL Chaotic learning MAE Modified auto-encoderCNN Convolutional neural network MLP Multi-layered perceptronDCNN Deep CNN Mol2Vec Molecular to vectorsDNN Deep neural network R-CNN Regional convolutional networkDT Decision tree PNN Polynomial neural networks [141]FCN Fully convolutional network PR Polynomial regression [44]GA Genetic algorithm RF Random forest [22]GAE Generative auto-encoder RL Reinforcement learningGANs Generative adversarial nets RNN Recurrent neural networkGAN Generative adversarial network SVM Support vector machineGRU Gate recurrent unit SVR Support vector regression [13]KNN K -nearest neighbor TL Transfer learningLR Logistic regression t-SNE t-distributed stochastic neighborembedding [125]LSTM Long short-term memory [66] VAE Variational auto-encoder [186] serological diagnosis, chest X-ray and CT image inspection, and other noninvasive methods. Benefitting from the advantages of high sensitivity and specificity, real-time Reverse TranscriptasePolymerase Chain Reaction (RT-PCR) is the current standard detection technology in diagnosing theSARS-CoV-2 virus and bacterial infections. Using RT-PCR, 9 RNA positives were detected frompharyngeal swabs of patients, indicating that the SARS-CoV-2 virus had spread in communities ofWuhan, China, in early January 2020 [106]. The shedding of the SARS-CoV-2 virus detected in thethroat, lungs, and feces suggests multiple routes of virus transmission [217, 227]. However, RT-PCR facesthe limitations of a complicated sample preparation, low detection efficiency, and high false-negativerate [106, 212, 223]. 3/36sothermal nucleic acid amplification and blood testing methods are also commonly used for the rapidscreening of SARS-CoV-2 [103, 122, 223]. An ML classification method was used for blood testing toextract important routine hematological and biochemical characteristics and to provide COVID-19classification. In [223], 105 blood test reports were collected, of which, 27 were collected as positivesaples from patients with confirmed COVID-19, and for comparison, negative samples were collectedfrom patients with ordinary pneumonia, tuberculosis, and lung cancer. Each sample contains 49 featurevariables, including routine hematological and biochemical parameters. Next, the authors implementedthe RF algorithm [22] on the training samples to perform feature learning and classification. Based onthe extracted 11 key feature variables, they built an RF classifier and tested 253 samples of 169 patientswith suspected COVID-19 with an accuracy of 96.97%. Although AI technologies rarely participatedirectly in RT-PCR and blood testing, the viral load and COVID-19 case data collected in thesemethods provide important data sources for the subsequent AI-based analysis.
Medical imaging inspection is another widely used clinical approach for COVID-19 detection anddiagnosis. COVID-19 medical image inspection mainly includes chest X-ray and lung CT imaging. AItechnology plays an important role in medical image inspection and has achieved significant results inimage acquisition, organ recognition, infection region segmentation, and disease classification. It not onlygreatly shortens the time of a radiologist’s image diagnosis but also improves the accuracy andperformance of the diagnosis. We will discuss in detail the contributions of AI methods to chest X-rayand lung CT imaging.
CT imaging provides an important basis for the early diagnosis of COVID-19. The CT imagingmanifestations of COVID-19 are mainly Ground Glass Opacity (GGO) in the periphery of the subpleuralregion, and some are consolidated. If the situation improves, the area will be absorbed and form fibrousstripes. Examples of lung CT images of normal and COVID-19 cases are shown in Fig. 2 [42, 161, 184].
Figure 2.
Examples of lung CT images of normal and COVID-19 cases.The progress of CT image inspection based on AI usually includes the following steps: Region OfInterest (ROI) segmentation, lung tissue feature extraction, candidate infection region detection, andCOVID-19 classification. The representative AI architecture for COVID-19 CT image classification isshown in Fig. 3.The segmentation of lung organs and ROIs is a foundational step in AI-based image inspection. Itdepicts the ROIs in lung CT images (such as lungs, lung lobes, bronchopulmonary segments, andinfected regions or lesions) for further evaluation and quantification. Different DL models (such as4/36 igure 3.
Representative architecture of AI-based CT image classification and COVID-19 inspection.U-Net, V-Net, and VB-Net) have been used for CT image segmentation [32, 94, 114, 193, 226]. In [179],Shan et al . collected 549 CT images from patients with confirmed COVID-19 and proposed an improvedsegmentation model (named VB-NET) based on the V-NET [117] and ResNet [69] models. In [32], Chen et al . built a DL model based on the U-Net++ structure [244] to extract the ROIs from each CT imageand detect the training curve of suspicious lesions. In [226], Xu et al . used a 3D DL model to segmentthe infection regions from lung CT images. They then built a classification model using ResNet andlocation-attention structures and divided the segmented area images into three categories, such asCOVID-19, influenza-A viral pneumonia, and normal. In [114], Li et al . used the U-Net segmentationmodel to extract the lung organ from each lung CT image as an ROI. In [94], Jin et al . proposed anAI-based COVID-19 diagnostic system, which consists of a lung segmentation module and a COVID-19diagnostic module. The lung segmentation module is implemented based on Deeplabv1 [33]. In [193],Tang et al . used the VB-Net model [179] to accurately segment 18 lung regions and infected regions fromlung CT images and further calculated 63 quantitative features.Focusing on the detection and localization of candidate infection regions, different AI methods wereproposed in [65, 85, 181, 212]. In [65], Gozes et al . used commercial software to identify lung nodules andsmall opacities within the 3D lung volume. Then, they constructed a DL model consisting of U-Net andResNet structures, where the U-Net module was used to extract the ROI regions, and the ReseNet modelwas used to detect and classify diffuse turbidity and ground glass infiltration. The authors compared theCT images of 56 patients with confirmed COVID-19 and 101 noncoronavirus patients and analyzed theCT features of COVID-19 in detail. In [181], Shi et al . used a CNN model based on V-Net to segmentlung organs and infection regions from lung CT images. Then, they used the LASSO method tocalculate the best CT morphological features. Finally, the severity of COVID-19 was predicted andevaluated based on the best CT morphology and clinical features. In [212], Wang et al . collected 195 CTimages from 44 patients with COVID-19 and 258 CT images from 55 negative patients. They used theCNN model with the Inception structure [191] to classify randomly selected ROI images and predictCOVID-19 disease. In [85], Huang et al . used an InferReadTM CT pneumonia tool based on AI toquantitatively evaluate changes in the lung burden of patients with COVID-19. The tool includes threemodules: lung and lobe region extraction, pneumonia segmentation, and quantitative analysis. The CTimage features of COVID-19 pneumonia are divided into four types: mild, moderate, severe, and critical.Based on ROI segmentation and candidate infection region detection, the important features of ROIsand infection regions are extracted for COVID-19 classification [156]. In [156], Qi et al . collected 71 CTimages from 52 patients with confirmed COVID-19 in 5 hospitals. They used the pyradiomics method toextract 1,218 features from each CT image and then performed LR and RF methods on these features todistinguish between short-term and long-term hospital stay. In [180], Shi et al . used the VB-NETmodel [179] to segment the infection and lung fields from CT images and divided them based on 96features, including 26 volume features, 31 digital features, 32 histogram features, and 7 surface features.5/36ext, they proposed an iSARF method to classify features and predict COVID-19 disease. Comparativeexperiments showed that the iSARF method is superior to the LR, SVM, and NN methods. In [241],Zheng et al . proposed a 3D DCNN model (named DeCoVNet) to detect COVID-19 from CT images.The proposed DeCoVNet model includes three components. The first component uses vanilla 3Dconvolutional layers to extract lung image features, the second component consists of two 3D residualblocks that perform element conversion on the 3D feature maps, and the third component graduallyextracts the information in the 3D feature map through 3D max-pooling and outputs the probability ofCOVID-19. In [187], Song et al . collected 1990 CT images, including 777 images from 88 patients withCOVID-19, 505 images from 100 patients with bacterial pneumonia, and 708 images from 86 healthypeople. They proposed a DRE-NET DL model based on the pretrained ResNet50 structure andfunctional pyramid networks. DRE-NET extracts the top- K lesion features from each CT image topredict the classification of patients with COVID-19.The lack of large-scale datasets is the main challenge that hinders the implementation of AI-basedCT image inspection and affects diagnostic performance. To address these challenges, strategies such astransfer learning, data augmentation, and “Human-In-The-Loop” were used in [94, 179, 239]. In [94], Jin et al . used the ImageNet7 dataset [47] to pretrain the proposed 2D classification network. In [239], Zhao et al . provided a public COVID-19 CT scan dataset, including 275 COVID-19 cases and 195non-COVID-19 cases. They used data augmentation and TL methods to alleviate the shortage oftraining data. In terms of data augmentation, they used transformation operations to expand thetraining dataset, such as random transformation, cropping, and rotation. In terms of TL, theypretrained the DenseNet model [84] on the chest X-ray dataset [213] and then used the pretrained modelto predict COVID-19. In addition, a “Human-In-The-Loop” strategy was adopted to reduce theworkload of radiologists in annotating the training samples [179]. Radiologists annotate a small portionof training samples in the first batch of training. Then, they manually correct the segmentation resultsin the second batch and used them as annotations of the images. Iterative training is performed in thisway to complete the annotation of all training samples.It is commendable that several works provided open-source code of the designed models and onlineCOVID-19 CT image inspection systems. For example, Li [114], Jin [94], Zheng [241], and Zhao [239]published the proposed DL models on GitHub [88]. In addition, Song et al . [187] provided an online CTdiagnosis service. Wang et al . [212] provided a public website for CT image uploading and testing.In [32], Chen et al . developed a public online CT diagnostic system, and anyone can upload CT imagesfor self-diagnosis. More detailed information about AI-based CT image segmentation and classificationmethods is provided in Table 2 and Fig. 4. Compared with CT images, chest X-ray (CXR) images are easier to obtain in clinical radiologyinspections. Although CXR image inspection is a typical imaging method used for COVID-19 diagnosis,it is generally considered to be less sensitive than CT image inspection. Some CXR images of patientswith early COVID-19 showed normal characteristics. The radiological signs of COVID-19 CXR imagesinclude airspace opacity, GGO, and later mergers. In addition, the distribution of bilateral, peripheral,and lower regions is mostly observed. Examples of CXR images of normal and COVID-19 cases areshown in Fig. 5 from [124, 160]. The CXR image inspection process based on AI techniques usuallyincludes steps such as data preprocessing, DL model training, and COVID-19 classification. Therepresentative AI architecture for COVID-19 CXR image inspection is shown in Fig. 6.Unlike CT images, CXR image segmentation is more challenging because the ribs are projected ontosoft tissues, which is confused with image contrast. In this way, most DL models focus on theclassification of the entire CXR image, while few works focus on segmenting ROIs and lung organs fromCXR images. In [67], Hassanien et al . used a classification method to identify and classify COVID-19 onlung X-ray images through a multilevel threshold and SVM. A multilevel image segmentation thresholdwas used to segment the lung organs from the background, and then the SVM module classified theinfected lungs from the uninfected lungs.Focusing on COVID-19 classification based on CXR images, several studies built AI-based 6/36 able 2.
AI-based CT image segmentation and classification methods for COVID-19 inspection.
Literature Datasources Datasize COVID-19cases AI methods ACC/AUC Sensitivity SpecificityChen [32] private 35,355 20,886 U-Net++ 95.24% 100% 94.0%Gozes [65] private 157 56 U-Net, ResNet 99.6% 98.2% 92.2%Huang [85] private 842 842 U-Net - - -Jin [94] [11, 48] 2095 970 Deeplab, ResNet 94.98% 94.06% 95.47%Li [114] private 4,356 1,325 U-Net, ResNet 96.0% 90.0% 96.0%Qi [156] private 52 52 LR, RF 97.0% 100% 75.0%Shan [179] private 549 549 V-Net, ResNet - - -Shi [180] private 2685 1658 RF 87.9% 90.7% 83.3%Shi [181] private 196 196 V-Net 89.0% 82.2% 82.8%Song [187] private 1990 777 DRE-Net 94.0% 93.0% -Tang [193] private 176 176 RF 87.5% - -Wang [212] private 453 195 Inception 73.1% 74.0% 67.0%Xu [226] private 618 219 ResNet 86.7% - -Zhao [239] [239] 470 275 DenseNet 84.7% 76.2% -Zheng [241] private 630 630 DeCoVNet 90.1% 90.7% 91.1% http://121.40.75.149/znyx-ncov/index. https://github.com/ChenWWWeixiang/diagnosis covid19. https://github.com/bkong999/COVNet.git. http://biomed.nscc-gz.cn/server/Ncov2019. https://ai.nscc-tj.cn/thai/deploy/public/pneumonia ct. https://github.com/UCSD-AI4H/COVID-CT. https://github.com/sydney0zq/covid-19-detection. Figure 4.
Relationship between AI methods and applications of COVID-19 CT image inspection.classification models by nesting or combining existing ML and DL models. In [73], Hemdan et al .proposed a DL framework (called COVIDX-Net) to help radiologists automatically diagnose COVID-19based on CXR images. The proposed framework integrates 7 DCNN models with different structures,such as VGG19 [183], DenseNet201 [84], ResNetV2 [29], InceptionV3, InceptionResNetV2 [189],Xception [39], and MobileNetV2 [170]. Each model was separately trained on CXR images to classify thepatient’s status as COVID-19 positive or negative. In [178], Sethy et al . used different DCNN models inan SVM classifier to diagnose COVID-19 on CXR images. Eleven DCNN models were used as imagefeature extractors, such as AlexNet, GoogleNet, DenseNet, Inception, ResNet, VGG, XceptionNet, and7/36 igure 5.
Examples of chest X-ray images of normal and COVID-19 cases.
Figure 6.
Representative architecture of AI-based CXR image classification and COVID-19 inspection.InceptionResNet. Then, an SVM classifier was used to classify the extracted image features to determinewhether the input image was COVID-19. In [26], Castiglioni et al . collected CXR data (610 images,including 324 COVID-19 cases) from Lombardy, Italy, and constructed a DCNN model to predictCOVID-19 disease. The DCNN model is equipped with 10 CNN models, each of which uses theResNet-50 structure and has been pretrained on the ImageNet dataset. Ioannis et al . [9] evaluated theperformance of 5 DCNN models recommended for medical image classification, including VGG19,MobileNetV2, Inception, Xception, and InceptionResNetV2. To address the shortcomings of thesmall-scale COVID-19 dataset, each model was pretrained on the ImageNet dataset by using the TLstrategy. Then, each pretrained model was executed on the CXR dataset. In [137], Narin et al . also used3 typical DCNN models (ResNet50, InceptionV3, and InceptionResNetV2) to classify COVID-19 from asmall-scale CXR dataset (100 images, including 50 COVID-19 cases).Another group of works constructed specific DL models for COVID-19 classification [237]. Forexample, Zhang et al . [237] proposed a new DL model, which consists of a backbone network, aclassification module, and an anomaly detection module. The backbone network extracts the features ofeach input CXR image. The classification module and anomaly detection module use the extracted8/36eatures to generate classification scores and scalar anomaly scores, respectively. In [209], Wang et al .introduced a COVID-Net DCNN model to identify COVID-19 cases based on CXR images. TheCOVID-Net model uses a large number of convolutional layers in a projection-expansion-projectiondesign pattern. They collected 13,800 CXR images from 13,725 patients (including 183 COVID-19patients) to establish a CXR database (called COVIDx) for training COVID-Net. It is commendablethat the authors provided an open-source of the proposed model code and the COVIDx database.Similar to CT images, in CXR image inspection, there is also the problem of a lack of large-scaledatasets for DL model training. In [121], Loey et al . used the GAN model [86] to generate more CXRimages, thereby extending the scale of the CXR dataset. In addition, three DL models (Alexnet [109],GoogleNet [190], and ResNet18 [69]) were used to classify CXR images into four categories: COVID-19,normal, pneumonia bacteria, and pneumonia virus. In [127], Maghdid et al . used CNN and AlexNetmodels to train CXR and CT images to diagnose COVID-19 cases, respectively. Among them, theAlexNet model was pretrained on the ImageNet dataset to perform COVID-19 classification on thedatasets in [41, 124, 140]. Unlike existing TL and image augmentation methods, Afshar et al . designed acapsule network model (named COVID-CAPS) suitable for small-scale CXR datasets [2]. Each layer ofthe COVID-CAPS model contains multiple capsules, and each capsule represents a specific imageinstance at a specific position through multiple neurons. The capsule module [75] uses protocol routingto capture alternative models of spatial information and attempts to reach a consensus on the existenceof objects. In this way, the protocol uses information from instances and objects to identify therelationship between them without the need for large-scale datasets. More detailed information aboutAI-based CXR image classification methods for COVID-19 inspection is shown in Table 3 and Fig. 7.
Table 3.
AI-based CXR image classification methods for COVID-19 inspection.
Literature Data sources Datasize COVID-19cases AI methods AccuracyAfshar [2] [41, 138] 100 50 COVID-CAPS 95.7%Castiglioni [26] private 610 324 ResNet 89.0%Ioannis [9] [41, 102, 112] 1427 224 VGG, MobileNet, Inception, Xception, In-ceptionResNet 96.78%Hassanien [67] private 40 25 SVM 97.48%Hemdan [73] [41, 159] 50 25 VGG, DenseNet, ResNet, Inception, In-ceptionResNet, Xception, MobileNet 80.0%Loey [121] [41, 138] 306 69 Alexnet, GoogleNet, ResNet 80.6%Maghdid [127] [41, 124, 140] 170 60 CNN, AlexNet 94.0%Narin [137] [41, 138] 100 50 ResNet, Inception, InceptionResNet 98.0%Sethy [178] [41, 138] 50 25 AlexNet, DenseNet, GoogleNet, Inception,ResNet, VGG, Xception, InceptionResNet,SVM 95.38%Wang [209] [24, 41, 138] 13,800 183 COVID-Net 92.6%Zhang [237] [41, 213] 1531 100 new DL 96.0% https://github.com/lindawangg/COVID-Net. https://github.com/ShahinSHH/COVID-CAPS. In addition to RT-PCR detection and image inspection techniques, some noninvasive measurementmethods have also been used for COVID-19 detection and diagnosis, including cough sound judgmentand breathing pattern detection.(1) Monitoring COVID-19 through AI-based cough sound analysis.Schuller et al . [174] discussed the potential application of computer audition (CA) and AI in theanalysis of cough sounds in patients with COVID-19. They first analyzed the CA’s ability toautomatically recognize and monitor speech and cough under different semantics, such as breathing, dry9/36 igure 7.
Relationship among CXR datasets, AI methods, and applications of CXR image inspection.and wet coughing or sneezing, speech during colds, eating behaviors, drowsiness, or pain. Then, theysuggested applying the CA technology to the diagnosis and treatment of patients with COVID-19.However, due to the lack of available datasets and annotation information, there is no report on theapplication of this technology in COVID-19 diagnosis. Similarly, Iqbal et al . [90] also discussed anabstract framework that uses the speech recognition function of mobile applications to capture andanalyze the cough sounds of suspicious persons to determine whether the user is healthy or suffers from arespiratory disease. In [214], Wang et al . analyzed the respiratory patterns of patients with COVID-19and other breathing patterns of patients with influenza and the common-cold. In addition, theyproposed a respiratory simulation model (named BI-AT-GRU) for COVID-19 diagnosis. TheBI-AT-GRU model includes a GRU neural network with a bidirectional and attention mechanism andcan classify 6 types of clinical respiratory patterns, such as Eupnea, Tachypnea, Bradypnea, Biots,Cheyne-Stokes, and Central-Apnea.(2) COVID-19 diagnosis based on noninvasive measurements.In [128], Maghdid et al . designed an abstract framework for COVID-19 diagnosis based on smartphone sensors. In the proposed framework, smart phones can be used to collect the diseasecharacteristics of potential patients. For example, the sensors can acquire the patient’s voice through therecording function and can obtain the patient’s body temperature through the fingerprint recognitionfunction. Then, the collected data are submitted to the AI-supported cloud server for disease diagnosisand analysis.
The virology and pathogenesis of SARS-CoV-2 are one of the most important scientific studies in thefields of biology and medicine. Scientists have analyzed the virus characteristics of SARS-CoV-2 throughproteomics and genomic studies [7, 77, 123]. In the field of virology, the origin and classification ofSARS-CoV-2, the physical and chemical properties, receptor interactions, cell entry, and the ecology andgenomic variation in SARS-CoV-2 have been studied [123, 217, 242]. We mainly discuss the contributionof AI in the pathological research of SARS-CoV-2 from the perspective of proteomics and genomics.10/36 .1 Proteomics
Since the advent of SARS-CoV-2, there have been a large number of research achievements in proteomics.Five types of structural proteins of SARS-CoV-2 were confirmed, including nucleocapsid (N) proteins,envelope (E) proteins, membrane (M) proteins, and spike (S) proteins [145, 206, 240]. In addition, otherproteins translated in the host cells essential for virus replication have also attracted the attention ofresearchers, such as non-structural protein 5 (NSP5) and 3C-like protease (3CLpro). Moreover, severalstudies have shown that SARS-CoV-2 uses the human Angiotensin-Converting Enzyme 2 (ACE2) toenter the host [77, 242]. In this field, AI techniques are used to predict protein structures and analyze theinteraction network between proteins and drugs. The representative AI architecture for protein structurepredication is shown in Fig. 8.
Figure 8.
Representative AI architecture for protein structure predication.In [176, 177], Senior et al . used DL models to implement the AlphaFold system for protein structureprediction. The AlphaFold system uses a ResNet model [69] to analyze the covariance and amino acidresidue contacts in homologous gene sequences and to predict the corresponding protein structures. TheAlphaFold system consists of a feature extraction module and a distance prediction neural network. Thefeature extraction module is responsible for searching for protein sequences that are similar to the inputprotein sequences and constructing the multiple sequence alignment (MSA). The module simultaneouslygenerates residual position and sequence contour features, and the output of 485 feature parameters areinput into the distance prediction neural network. The distance prediction neural network is atwo-dimensional (2D) ResNet structure, which is responsible for accurately predicting the distancebetween all residue pairs of every two protein sequences. The authors added a one-dimensional outputlayer to the network to predict the accessible surface area, distance map, and secondary structure of eachresidue. Finally, the generated potential is optimized by gradient descent to generate protein structures.Based on [176, 177], Jumper et al . [96] used the AlphaFold system to predict the structure ofSARS-CoV-2 membrane proteins. They published the predicted protein structures such as 3a, Nsp2,Nsp4, Nsp6, and papain-like proteases. Although the structure of these proteins has not been verified byclinical experiments, this publication allows researchers to quickly conduct SARA-CoV-2 studies.In [145], Ortega et al . used a computational method to detect changes in the S1 subunit of the spikereceptor-binding domain and determined mutations in the SARS-CoV-2 spike protein sequence, whichmay be beneficial for studying human-to-human transmission. They collected sequences for modelingand constructed the SARS-CoV-2 spike protein model from the Protein Data Bank (PDB) [19] and usedSWISS-MODEL software [12] to construct the SARS-CoV-2 spike protein model. Then, Z-docksoftware [151] was used to dock between the spike protein and ACE2, and a clustering algorithm wasused to cluster the docking results. The work indicated that the SARS-CoV-2 spike protein has a higheraffinity for human ACE2 receptors.Another branch of AI-assisted proteomics research involves finding new compounds and drugcandidates for the treatment of COVID-19 by building interactive networks and knowledge mapsbetween proteins and drugs. Please see Section 4 for details. 11/36 .2 Genomics
Genomics is mainly used in SARS-CoV-2 to analyze the origin of SARS-CoV-2, vaccine development,and PT-PCR detection. Various AI algorithms are applied for similarity comparisons of gene sequences,gene fragments, and miRNA prediction [46, 164]. In [164], Randhawa et al . used different ML methods toanalyze the pathogen sequences of COVID-19 and identified the inherent features of the viral genomes,thereby rapidly classifying new pathogens. They collected the complete reference genome of the COVIDvirus from NCBI [53], the bat β -coronavirus from GISAID [61], and all available virus sequences fromVirus-Host DB [135]. Each genomic sequence was mapped to a corresponding genomic signal in adiscrete digital sequence by using chaotic game representation [93]. In addition, the amplitude spectrumof these genomic signals was calculated by using a discrete Fourier transform. On this basis, they used 6ML classification models to train the above sequence distance matrix and compared their performance.Finally, they conducted the trained ML models on 29 COVID-19 sequences to classify COVID-19pathogens. The results of this work support the hypothesis that COVID-19 originated in bats and itsclassification as a β -coronavirus.In [46], Demirci et al . performed a miRNA prediction on the SARS-CoV-2 genome based on 3 MLmethods and identified miRNA-like hairpins and microRNA-mediated SARS-CoV-2 infectioninteractions. They collected the complete COVID-19 genome from NCBI [53] and human-mature miRNAsequences from miRBase [108]. The genomic sequences are transcribed and divided into multipleoverlapping fragments, which are folded into a secondary structure to extract the hairpin structure. Onthis basis, the authors used 3 ML methods (e.g., DT, Naive Bayes, and RF) to predict the category ofeach hairpin and determined the similarity between the hairpins and human miRNA. They searched formature miRNA targets in human and SARS-CoV-2 genes and analyzed the potential interactionsbetween SARS-CoV-2 miRNAs and human genes and between human miRNAs and SARS-CoV-2 genes.Finally, the gene ontology of SARS-CoV-2 miRNA targets in human genes were analyzed, and thesimilarity between SARS-CoV-2 miRNA candidates and mature miRNAs of any known organism wasevaluated using the PANTHER classification system [134].In [133], Metsky et al . used genomic and AI technologies to rapidly design nucleic acid detectionassays and improved current RT-PCR testing of SARS-CoV-2. They developed a CRISPR tool that usesenzymes to edit the genome by cutting specific genetic code chains and used different ML methods topredict the diversity of the target genome. The authors designed the RT-PCR test method through theCRISPR tool, and it can effectively detect 67 respiratory viruses, including SARS-CoV-2. Based on proteomics and genomics research, a variety of drug and vaccine development programs havebeen proposed for SARS-CoV-2 and COVID-19. The application of AI in the development of new drugsand vaccines is one of the main contributions in smart medicine and plays an important role in the battleagainst COVID-19.
In the field of drug development, AI technologies can screen existing drug candidates for COVID-19 byanalyzing the interaction between existing drugs and COVID-19 protein targets. In addition, AItechnologies can help to discover new drug-like compounds against COVID-19 by constructing newmolecular structures that have inhibitory effects on proteases at the molecular level. The representativeAI architecture for new drug-like compound discover is shown in Fig. 9.Drug development can be divided into small-molecule drug discovery and biological productdevelopment. Small-molecule drug discovery mainly focuses on chemically synthesized small molecules ofactive substances, which can be made into small-molecule drugs through chemical reactions betweendifferent organic and inorganic compounds. One group of AI-based drug development focuses on thediscovery of new drug-like compounds at the molecular level. In [18, 182], Beck et al . proposed a 12/36 igure 9.
Representative AI architecture for new drug-like compound discover.DL-based drug-target interaction model (MT-DTI) to predict potential drug candidates for COVID-19.The MT-DTI model uses SMILES strings and amino-acid sequences to predict target proteins with 3Dcrystal structures. The authors collected the amino-acid sequences of 3C-like proteases and relatedantiviral drugs and drug targets from the databases of NCBI [53], Drug Target Common (DTC) [194],and BindingDB [120]. In addition, they used a molecular docking and virtual screening tool (AutoDockVina [202]) to predict the binding affinity between 3,410 drugs and SARS-CoV-2 3CLpro. Theexperimental results provided 6 potential drugs, such as Remdesivir, Atazanavir, Efavirenz, Ritonavir,Dolutegravir, Kaletra (lopinavir/ritonavir). Note that Remdesivir shows promising in clinical trial.In [136], Moskal et al . used AI methods to analyze the molecular similarity between anti-COVID-19drugs (termed “parents”) and drugs involving similar indications to screen out second-generation drugs(termed “progeny”) for COVID-19. They first used the Mol2Vec [91] method to convert the molecularstructure of the parent drugs into a high-dimensional vector space, treated the drug molecule as a“sentence”, and mapped its molecular substructure to a “word”. Then, they used the VAE [186] model togenerate SMILES strings with similar 3D shape and pharmacodynamic properties to a given seedmolecule [63]. In addition, CNN, LSTM, and MLP models are used to generate the correspondingSMILES strings and molecules. The authors selected 71 parent drugs as seed molecules from theliterature and selected 4456 drugs as candidate progeny drugs from ZINC [245] and ChEMBL [56].In [23], Bung et al . committed to the development of new chemical entities for the SARS-CoV-23CLpro based on DL technology. They constructed an RL-based RNN model to classify proteaseinhibitor molecules and obtained a smaller subset that favored the chemical space. Then, they collected2515 protease inhibitor molecules in SMILES format from the ChEMBL database as training data, whereeach SMILES string is regarded as a time series, and each position or symbol is regarded as a time point.The output of small molecules was docked to the 3CLpro structure with minimal energy and rankedbased on the virtual screening score obtained by selecting candidates of anti-SARS-CoV-2 [202]. In [192],Tang et al . analyzed 3CLpro with a 3D structure similar to SARS-CoV and evaluated it as an attractivetarget for anti-COVID-19 drug development. They proposed an advanced deep-Q learning network(called ADQN-FBDD) to generate potential lead compounds of SARS-CoV-2 3CLpro. They collected284 reported molecules as SARS-CoV-2 3CLpro inhibitors. These molecules were split using theimproved BRICS algorithm [45] to obtain the target fragment library of SARS-CoV-2 3CLpro. Then, theproposed ADQN-FBDD model trains each target fragment and predicts the corresponding molecules andlead compounds. Through the proposed Structure-Based Optimization Policy (SBOP), they finallyobtained 47 derivatives with inhibitory effects on SARS-CoV-2 3CLpro from these lead compounds,which are regarded as potential anti-SARS-CoV-2 drugs.Another group of studies focused on screening candidate biological products for COVID-19.Biological products are a type of protein products with therapeutic effects, which are mainly combinedwith specific cell receptors involved in the disease process. Biological products are prepared frommicrobial cells such as genetically modified bacteria, yeast, or mammalian cell strains throughbiotechnology processes. In [79], Hu et al . established a multitask DL model to predict the possiblebinding between potential drugs and SARS-CoV-2 protein targets, thereby selecting available drugs forSARS-CoV-2. They first collected 8 SARS-CoV-2 viral proteins from GHDDI [58] as potential targets.13/36he proposed DL model is based on the AtomNet model [188, 205] and includes a shared layer to learnthe joint representation of all tasks and a task processing layer for performing specific tasks. Byfine-tuning the DL model using a coronavirus-specific dataset, the model can predict the possible bindingbetween the drugs and the protein targets and output the binding affinity score. Based on existingstudies, RdRp, 3CLpro, and papain-like protease have been confirmed as the three principal targets ofSARS-CoV-2 [60, 146, 206]. Based on the prediction results [113, 210], the authors selected the top 10potential drugs with a high likelihood of inhibition for each target. In [98], Kadioglu et al . usedHigh-Performance Computing (HPC), virtual drug screening, molecular docking, and ML technologies toidentify SARS-CoV-2 drug candidates. After performing virtual drug screening and molecular docking,two supervised ML models(e.g., NN and Naivebayes) were used to analyze clinical drugs and testcompounds to construct corresponding drug likelihood prediction models. Several approved drugs,including those used for the hepatitis C virus (HCV), the enveloped ssRNA virus, and other infectiousdiseases, were selected as SARS-CoV-2 drug candidates.Facing the known COVID-19 protease target 3CLpro, Zhavoronkov et al . [240] designed asmall-molecule drug-discovery pipeline to produce 3CLpro inhibitors, used 3CLpro’s crystal structure,homology modeling, and co-crystallized fragments to generate 3CLpro molecules. They collected thecrystal structure of COVID-19 3CLpro from [230] and constructed a homology model. At the same time,molecules with activity on various proteases were extracted from [56, 89] and constituted a proteasepeptidomimetic dataset with 5,891 compounds. Then, they used 28 ML methods (such as GAE, GAN,and GA) and RL strategies to separately train input datasets (e.g., crystal structure, homology model,and co-crystal ligands), and generated new molecular structures with a high score. In [78], Hofmarcher etal . used a ChemAI DL model [154] based on the SmilesLSTM structure [76] to test the resistance of themolecules to COVID-19 proteases. They collected 3.6 million molecules from ChEMBL [56], ZINC [245],and PubChem [104] and formed a training dataset. Then, the ChemAI model was trained on the datasetin a multitask parallel training way, where the output neurons of the model represent the biologicaleffects of the input molecules. The authors used the ChemAI model to predict the inhibitory effects ofthese molecules on the 3CLpro and PLpro proteases of COVID-19. These molecules have a binding,inhibitory, and toxic effect on the targets. A list of COVID-19 drug development methods based on AI isprovided in Table 4.
Currently, there are 3 types of COVID-19 vaccine candidates, such as (1) whole virus vaccines, (2)recombinant protein subunit vaccines, and (3) nucleic acid vaccines [38, 238]. AI technology has beeninvolved in the design and development of COVID-19 vaccines. Compared with explicit applications inother fields, AI technology is usually used in the sub-processes of vaccine development in an implicitmanner.The AI algorithms of netMHC and netMHCpan are used in the development of COVID-19 vaccinesfor epitope prediction [74, 97, 215]. In [74], Herst et al . obtained the SARS-CoV-2 protein sequences fromGenBank and used the MSA algorithm to trim the nucleocapsid phosphoprotein sequences to possiblepeptide sequences. On this basis, they used netMHC and netMHCpan AI algorithms to train and predictpeptide sequences [8, 97]. The pan variant of netMHC integrates the in-vitro objects of 215 HLAs forprediction. Finally, they used the average value of the ANN, SVM, netMHC and netMHCpan methodsto calculate the vaccine candidates. In [215], Ward et al downloaded the SARS-CoV-2 nucleotidesequences from the NCBI [53] and GISAID [61] databases, and generated a consensus sequence for eachSARS-CoV-2 protein. The sequences can be used as references for prediction, specificity, and epitopemapping analysis. Next, the authors used different epitope prediction tools to predict B cell epitopes andmap them to the amino acid sequences of each gene. On this basis, they used the AI-based netMHCpanalgorithm to predict HLA-1 peptides and obtained a total of 2,915 alleles in all peptide lengths.BLASTp tool [6] was used to locate the short amino acid epitope sequences to the canonical sequences ofSARS-CoV-2 proteins. Finally, the author provided an online tool that provides functions ofSARS-CoV-2 genetic variation analysis, epitope prediction, coronavirus homology analysis, andcandidate proteome analysis. 14/36 able 4.
Drug development of COVID-19 based on AI methods.
Literature AI methods Role of AI methods COVID-19targets PotentialdrugsTang [192] RL, DQN predict molecules and leadcompounds for each targetfragment 3CLpro 47 compoundsZhavoronkov [240]
28 ML generate new molecularstructures for 3CLpro 3CLpro 100 moleculesBung [23] RNN, RL classify protease inhibitormolecules 3CLpro 31 compoundsHofmarcher [78] ChemAI predict inhibitory effectsof molecules on COVID-19proteases 3CLpro, PLpro 30,000moleculesKadioglu [98] NN, Naivebayes construct drug likelihoodprediction model spike protein, · · · · · ·
10 drugsBeck [18, 182] MT-DTI predict binding affinity be-tween drugs and proteintargets 3CLpro, RdRp,helicase, · · · https://github.com/tbwxmu/2019-nCov. https://github.com/ml-jku/sars-cov-inhibitors-chemai. In [143], Ong et al . used ML and Reverse Vaccinology (RV) methods to predict and evaluatepotential vaccines for COVID-19. They used RV to analyze the bioinformatics of pathogen genomes toidentify promising vaccine candidates. They obtained the SARS-CoV-2 sequences and all proteins of the6 known human coronavirus strains from the NCBI [53] and UniProt [17] databases. Then, they usedVaxign and Vaxign-ML [71, 142] to analyze the complete proteome of the coronaviruses and predictedtheir service biological characteristics. Next, they improved the Vaxign-ML model based on ML and RVusing LR, SVM, KNN, RF, and XGBoost methods and predicted the protein level of all SARS-CoV-2proteins. The nsp3 protein was selected for phylogenetic analysis, and the immunogenicity of nsp3 wasevaluated by predicting T cell MHC-I and MHC-II and linear B cell epitopes.In [158], Qiao et al . used DL to predict the patient’s mutated new antigen and identified the bestT-cell epitope for peptide-based COVID-19 vaccines. They first sequenced the diseased cells in thepatient’s blood and extracted 6 human leukocyte antigen (HLA) types and T-cell receptor (TCR)sequences. Then, they proposed the DeepNovo model to train the patient’s immune peptide and toidentify the best T-cell epitope set based on a person’s HLA alleles and immune peptide groupinformation. The DeepNovo model uses LSTM and RNN structures to capture sequence patterns inpeptides or proteins and predicts HLA peptides from conserved regions of the virus, thereby predictingnew mutant antigens in patients. In addition, they used the IEDB [204] tool to predict theimmunogenicity of 177 peptides. They suggested designing an epitope-based COVID-19 vaccinespecifically for each person based on their HLA alleles.The prediction of immune stimulation ability is an important part of vaccine designing [162, 166].Different ML methods and position-specific scoring matrices (PSSM) are usually used to predict epitopeand immune interactions, thereby predicting the generation of adaptive immunity in the target host.In [162], Rahman et al . used immuno-informatics and comparative genomic methods to design amulti-epitope peptide vaccine against SARS-CoV-2, which combines the epitopes of S, M, and E proteins.They used the Ellipro antibody epitope prediction tool [87] to predict linear B cell epitopes on the Sprotein. Ellipro uses multiple ML methods to predict and visualize a given protein sequence or B-cellepitope in the structure. In addition, Sarkar et al . [172] studied the epitope-based vaccine design for15/36OVID-19 and used the SVM method to predict the toxicity of the selected epitopes. In [153], Prachar et al . used 19 epitope-HLA combined prediction tools including IEDB, ANN, and PSSM algorithms topredict and verify 174 SARS-CoV-2 epitopes.
Thanks to the developed information and multimedia technology, the outbreak and spread of COVID-19were reported in a timely and accurate manner. The number of suspected, confirmed, cured, and deadCOVID-19 cases in each country/region is announced in real time. In addition, passenger traveltrajectories and related big data are shared for scientific research. Based on the rich data, numerousresearchers have participated in the prediction, spread, and tracking of the COVID-19 outbreak.
Researchers collected clinical COVID-19 case data and used different AI methods to extract importantfeatures and to predict the mortality and survival rate of patients with COVID-19. The representativeAI architecture for prediction of patient mortality and survival rate is shown in Fig. 10.
Figure 10.
Representative AI architecture for prediction of patient mortality and survival rate.In [152], Pourhomayoun et al . used 6 AI methods to predict the mortality rate of patients withCOVID-19. They used public data of patients with COVID-19 from 76 countries around the world [225],and counted 112 features, including 80 medical annotations and disease features and 32 features from thepatients’ demographic and physiological data. Based on the filtering method and wrapper method, 42best features were extracted, such as demographic features, general medical information, and patientsymptoms. On this basis, 6 AI methods (such as SVM, NN, RF, DT, LR, and KNN) are used to predictthe mortality of patients with COVID-19. In [173], Sarkar et al . used the RF model to analyze therecords of 433 patients with COVID-19 from Kaggle [43] and identified the important features and theirimpact on mortality. Experimental results show that patients over 62 years of age have a higher risk ofdeath. In [228, 229], Yan et al . analyzed a blood sample dataset of 404 patients with COVID-19 inWuhan, China, and used the XGBoost classification method [37] to select three important biomarkersand to predict individual patient survival rates. Experimental results with an accuracy of 90% indicatedthat higher LDH levels seem to play an important role in distinguishing the most critical COVID-19cases.
BlueDot [21] and Metabiota [132] are two AI companies that made accurate predictions for theCOVID-19 outbreak. BlueDot collected large-scale heterogeneous data from various sources, such asnews reports, global ticketing data, animal diseases, global infectious disease alerts, and real-time climateconditions. Then, it used filtering tools to narrow its focus; used various ML and Natural LanguageProcessing (NLP) techniques to detect, mark, and display the potential risk frequency of COVID-19; andpredicted the outbreak time of transmission. It is worth mentioning that 9 days before the official16/36nnouncement of the COVID-19 outbreak, BlueDot accurately predicted the epidemic of COVID-19 andcities with a high risk of virus outbreaks. Metabiota collected large-scale data from social and nonsocialsources (such as biology, socioeconomic, political, and environmental data) and used technologies such asAI, ML, big data, and NLP to accurately predict the outbreak, spread, and intervention measures ofCOVID-19. More AI-based COVID-19 outbreak and transmission prediction methods are shown in Table5.
Table 5.
COVID-19 outbreak and transmission prediction based on AI methods.
Literature Data sources Methods Country/regionHuang [82] Yang [231], WHO [216] CNN, LSTM, MLP, GRU ChinaHu [80, 81] The Paper [148], WHO [216] MAE, clustering ChinaYang [233] Baidu [16] SEIR, LSTM ChinaFong [51, 52] NHC [139] SVM, PNN ChinaAi [3] WHO [54, 216] ANFIS, FPA China, USARizk [168] WHO [216] ISACL-MFNN USA, Italy, SpainGiuliani [62] Italy [144] EMTMGL ItalyAyyoubzadeh [14] Worldometer [218], Google [201] LR, LSTM IranMarini [129, 130] Swiss population Enerpol SwitzerlandLai [110] IATA [126], Worldpop [219] ML GlobalPunn [155] JHU CSSE [49] SVR, PR, DNN, LSTM,RNN GlobalLampos [111] MediaCloud [131], PHE [64],ECDC [55] Transfer learning Global
Although the source of the COVID-19 epidemic has not yet been identified, it was first reported inWuhan, China. Therefore, the outbreak and spread of COVID-19 in China have received extensiveattention. In [82], Huang et al . used 4 DL models, such as CNN, LSTM, GRU, and MLP to train andpredict the COVID-19 case data from 7 severe epidemic cities in China. The input of these DL models isthe features of the COVID-19 cases, including the number of confirmed cases, cured cases, and deaths.Based on the input of the previous 5 days, each model can predict the number of COVID-19 cases forthe following few days. The architecture of the COVID-19 outbreak prediction model based on AImodels is shown in Fig. 11.
Figure 11.
Architecture of COVID-19 outbreak prediction model based on DL models.In [80, 81], Hu et al . used AI methods such as MAE and clustering algorithms to predict the numberof confirmed COVID-19 cases in different provinces and cities in China. In addition, they clustered 34provinces and cities in China into 9 clusters based on the prediction results and further predicted the17/36pread of COVID-19 among provinces and cities. In [233], Yang et al . used the SEIR model [101] andthe LSTM model to predict COVID-19 in China. The population migration data and the latestCOVID-19 epidemiological data from Baidu [16] were input into the SEIR model to derive the epidemiccurves. In addition, they used SARS data from 2003 to pretrain the LSTM model to predict COVID-19for the following few days, in which epidemiological parameters, such as the transmission, incubation,recovery probability, and the number of deaths, were selected as input features. Both the SEIR andLSTM models predicted a daily infection peak of 4,000 in the first week of February. In [51, 52], Fong etal . obtained early COVID-19 epidemiological data from NHC [139]. Then, they used traditional timeseries data analysis methods (e.g., ARIMA, Exponential, and Holt-Winters), ML methods (e.g., KR,SVM, and DT), and AI methods (e.g., PNN) to analyze and predict future outbreaks.In addition to China, the outbreak and spread of COVID-19 in other countries (including the UnitedStates, Italy, Spain, Iran, and Switzerland) have also received widespread attention. In [3], Ai et al .proposed an improved ANFIS method [92] to predict the number of COVID-19 cases. The proposedsystem connects fuzzy logic and neural networks and uses and enhanced Flower Pollination Algorithm(FPA) [232] for model parameter optimization and model training. In [168], Rizk et al . proposed animproved Multi-layer Feed-forward Neural Network (ISACL-MFNN) model, which uses an InternalSearch Algorithm (ISA) to optimize model parameters and uses the CL strategy to enhance the ISAperformance. From the official COVID-19 dataset reported by the WHO [216], data from January 22,2020, to April 3, 2020, in the United States, Italy, and Spain were collected to train the ISACL-MFNNmodel and to predict the confirmed cases within the next 10 days. In [62], Giuliani et al . collected thenumber of infected people in Italian provinces [144] and used the EMTMGL model to simulate andpredict the spatial and temporal distribution of COVID-19 infection in Italy. In [14], Ayyoubzadeh usedreal-time COVID-19 epidemic data from Google Trends [201] and Worldometer [218] to predictCOVID-19 cases in Iran. They collected daily epidemic data and saved them as a time series data formatand then used the LR and LSTM models to make predictions, thereby obtaining the outbreak andspread trend of COVID-19 in Iran. In [129, 130], Marini et al . developed an agent-based AI platform topredict the development of COVID-19 in Switzerland. The system accepts the entire Swiss population asinput data to simulate and predict the spread of COVID-19 in Switzerland. It simulates the people’sdaily trajectories by calibrating the micro-census data and effectively predicts the individual contactsand possible transmission routes.Many studies have likewise focused on the prediction of the spread of COVID-19 around the world.They collected a large amount of travel data, mobile phone data, and social media data and used AImethods to accurately predict the potential transmission range and transmission route of COVID-19.In [110], Lai et al . collected a large amount of travel and mobile phone data from [219] and constructedcorresponding models to predict the transmission risk of COVID-19 in different countries. On this basis,they established air travel network models between domestic cities and cities in other countries to predictrisk cities at home and abroad. In [155], Punn et al . used 2 ML models (e.g., SVR [13] and PR [44]) and3 DL regression models (e.g., DNN, LSTM [66], and RNN) to predict real-time COVID-19 cases.In [111], Lampos et al . used an automatic crawling tool to obtain daily confirmed COVID-19 case dataand related articles from online media such as MediaCloud [131], Public Health England (PHE) [64], andEuropean Centre for Disease Prevention and Control (ECDC) [55]. They used the TL strategy to transferthe COVID-19 model of the country where the disease spread to other countries that are still in the earlystage of the epidemic curve, and thus achieving the target country’s epidemic prediction. In addition,companies such as Microsoft Bing [20], Google [201], and Baidu [16] have aggregated multiple availabledata sources and developed COVID-19 global tracking systems to provide a visual tracking interface.In addition to AI methods, various methods based on statistics and epidemiology are used to predictthe outbreak and spread of COVID-19. In [70], He et al . collected the highest viral load in thepharyngeal swabs of 94 patients with confirmed COVID-19. They fitted a generalized additive modelwith identity links and smooth spline curves to analyze its overall trend. A gamma distribution wasfitted to the transmission pair data to evaluate the serial interval distribution. The results of statisticalanalysis showed that the patients with confirmed COVID-19 reach the peak of virus shedding before orduring symptom onset, and some kinds of transmission may occur before the initial symptoms. In [208],18/36ang et al . determined a set of technical indicators (e.g., number of infection cases in the hospital, dailyinfection rate, and daily cured rate) that reflect the infection status of COVID-19. Next, they proposed acalculation method based on statistical theory to quantify the iconic characteristics of each period andpredict the turning point in the development of the epidemic. In addition, numerous studies based onthe Susceptible-Infected-Recovered (SIR) and SEIR models have studied the spread of COVID-19 froman epidemiological perspective. Please see [25, 119, 169, 185, 195, 199, 234] for more information.
When COVID-19 appeared, most countries in the world adopted different forms of social control, socialalienation, school closures, and blockade measures to prevent the spread of the epidemic [203]. AItechnologies have been widely used in epidemic control and social management, including individualtemperature detection, video tracking, contact tracking, intelligent robots, etc. Many countries have usedsmart devices equipped with AI to detect suspicious persons in public transportation places such asairports and train stations [40, 167]. For example, infrared cameras are used to scan for hightemperatures in a crowd, and different AI methods perform efficient analysis to detect whether anindividual is wearing a mask in real time. In addition, DL-based video tracking technology is used todetect and track suspicious COVID-19 patients in public places [31]. Moreover, at the entrances andexits of cities, the identity information of each passing person was collected. Then, AI-based systems areused to efficiently query the travel history and trajectory of each passing individual to check whetherthey are from a region seriously affected by COVID-19 [35, 197].AI technologies are also used in contact tracking of patients with COVID-19 [100]. For each patientwith confirmed COVID-19, personal data such as mobile phone positioning data, consumption records,and travel records may be integrated to identify the potential transmission trajectory [57]. In addition,when people are in social isolation, mobile phone positioning and AI frameworks can assist thegovernment in better understanding the status of individuals [165]. Moreover, intelligent robots are usedto perform site disinfection and product transfer, and mobile phone positioning functions are used todetect and track the distribution and flow of personnel.Another group of studies focused on the impact of various social control strategies on the spread ofCOVID-19. In [72], Hellewell et al . proposed a random propagation model to analyze the success rate ofdifferent social control approaches to prevent the spread of COVID-19. By quantifying parameters suchas contact tracking and quarantined cases, the proposed model analyzed the maximum number of casestracked every week and measured the feasibility of public health work. In [105], Kissler et al . establisheda mathematical model to assess the impact of interventions on the prevalence of COVID-19 in theUnited States, and intermittent grooming measures were recommended to maintain effective control ofCOVID-19. Koo et al . [107] applied the population and personal behavior data of Singapore to theinfluenza epidemic simulation model and assessed different social isolation strategies on the dynamictransmission of COVID-19. Similarly, Chang et al . [28] constructed a simulation model to evaluate thepropagation and control of COVID-19 in Australia. The model analyzes the propagation characteristicsof COVID-19 and the impact of different control strategies on the propagation results.
The implementation and performance improvement of AI greatly depends on the large-scale availabledata and resources. Therefore, we compiled available public resources that can be used for COVID-19disease diagnosis, virology research, drug and vaccine development, and epidemic and transmissionprediction. Three types of data and resources were summarized, including medical images, biologicaldata, and informatics resources. 19/36 .1 Medical Images
We collected 17 groups of COVID-19 medical images such as CXR and CT images from individualresearchers and organizations. Among them, the CXR image data set published by Cohen et al . [41] iswidely cited, which is a collection of CXR images from multiple references. In addition, many researchersuploaded CXR and CT images to Kaggle [15, 112, 124, 138, 160] for COVID-19 research. Moreover,organizations such as the British Society of Thoracic Imaging (BSTI), Eurorad, and Radiopaedia alsoreleased online CXR and CT images. Table 6 displays the detailed description of medical image dataresources of COVID-19.
Table 6.
Medical image data resources for COVID-19 research.
Data sources Data type Cited by Refs.Zhao [239] CT images [239]HRCT [48] CT images [94]Armato [11] CT images [94]Coronacases [42] CT images -Medical segmentation [175] CT images -Cohen [41] CXR images [2, 9, 73, 121, 127, 137, 178, 209, 237]Wang [213] CXR images [237, 239]COVIDx [24] CXR images [209]Adrian [159] CXR images [73]COVID-Net [209] CXR images [209]Kermany [102] CXR images [9]Mendeley data [5] CXR images -Kaggle [15, 112, 124, 138, 160] CXR and CT images [2, 9, 121, 127, 137, 173, 178, 209]BSTI [140] CXR and CT images [127]SIRM [184] CXR and CT images -Eurorad [50] CXR and CT images -Radiopaedia [161] CXR and CT images -
We collected 10 biological data resources, such as NCBI, Protein Data Bank (PDB), UniProt, ClarivateAnalytics Integrity (CAI), Drug Target Common (DTC), and Virus-Host DB (VHDB), as shown inTable 7. These data resources provide abundant biological data resources, including gene sequences,proteins, drug molecules and compounds, and miRNA sequences.
Informatics resources such as COVID-19 situation reports, dashboards, COVID-19 cases, anddemographic data are gathered in Table 8. Among them, the World Health Organization (WHO), theNational Health Commission of the People’s Republic of China (NHC), the Disease Control andPrevention (CDC) provided real-time COVID-19 reports. Companies such as Baidu, Microsoft Bing,Google, and JHU CSSE provided online dashboards for COVID-19 spread tracking. Organizations suchas Govuk, Il Sole, and ECDC provided real-time COVID-19 cases in the UK, Italy, and Europe.
Although AI technologies have been used in fighting against COVID-19, and many studies have beenpublished, we note that the applications and contributions of AI in this work are still relatively limited.20/36 able 7.
Biological data sources for COVID-19 research.
Data sources Data type Description Cited by Refs.NCBI [53] Genome sequences Genome sequencing data of SARS-CoV-2 [18, 46, 143,164, 182]GISAID [61] Genome sequences Bat Betacoronavirus RaTG13 [164]VHDB [135] Genome sequences Virus sequences [164]PDB [149] Proteins 3D shapes of proteins, nucleic acids, andassemblies [19, 145]UniProt [17] Proteins SARS-CoV-2 protein entries and receptors [143]miRBase [108] miRNA sequences Human mature miRNA sequences [46]ZINC [245] Drug compounds drug compounds and molecules [78, 98, 136]DTC [194] Drug molecules Drug molecules for 3C-like proteases [18, 182]CAI [89] Drug discovery Empowering knowledge-based drug discov-ery and development [240]BindingDB [120] Amino-acid sequences Amino-acid sequences of 3C-like proteases [18, 182]
Table 8.
Informatics resources for COVID-19 research.
Data sources Data type Description Cited by Refs.WHO [216] Report COVID-2019 situation reports [3, 80, 82, 168]NHC [139] Report Real-time COVID-19 Report [51, 52]CDC [54] Report Weekly U.S. influenza surveillance report [3, 111]GitHub [88] Code Codes and datasets for COVID-19 study [24, 58, 239]Baidu [16] Dashboard Dashboard for COVID-19 tracking [233]Bing [20] Dashboard Dashboard for COVID-19 tracking -Google [201] Dashboard Dashboard for COVID-19 tracking [14]JHU CSSE [49] Dashboard Dashboard for COVID-19 tracking [155]Worldometer [218] COVID-19 cases COVID-19 pandemic cases [14]Govuk [64] COVID-19 cases COVID-19 cases in the UK [111]ECDC [55] COVID-19 cases COVID-19 cases worldwide [111]Il Sole [144] COVID-19 cases COVID-19 datasets in Italy [62]Worldpop [219] Demographic data Spatial demographic and air travel data [110]GHDDI [58] Community Drug discovery community [79]Humdata [68] Community community perceptions of COVID-19 -
We summarize the main challenges currently faced by AI against COVID-19 and provide thecorresponding suggestions.
At present, the applications of AI in COVID-19 research mainly faces four challenges: the lack ofavailable large-scale training data, massive noisy data and rumors, the limited knowledge on theintersection of computer science and medicine, and data privacy and human rights protection. • Lack of available large-scale training data. Most AI methods rely on large-scale annotated trainingdata, including medical images and various biological data. However, due to the rapid outbreak ofCOVID-19, there are insufficient datasets available for AI. In addition, annotating training samplesis very time-consuming and requires professional medical personnel. • Massive noisy data and rumors. Challenges arise from relying on the developed mobile Internetand social media; massive noise information and fake news about COVID-19 has been published onvarious online media without rigorous review. However, AI algorithms seem to be powerless in21/36udging and filtering the noise and erroneous data. This problem limits the application andperformance of AI, especially in epidemic prediction and transmission analysis. • Limited knowledge in the intersection of computer science and medicine. Many AI scientists arefrom computer science, but the application of AI in the COVID-19 battle requires in-depthcooperation in computer science, medical imaging, bioinformatics, virology, and many otherdisciplines. Therefore, it is crucial to coordinate the cooperative work of researchers from differentfields and integrate the knowledge of multiple subjects to jointly deal with COVID-19. • Data privacy and human rights protection. In the era of big data and AI, the cost of obtainingpersonal privacy data is very low. Faced with public health issues such as COVID-19, manygovernments want to obtain various types of personal information, including mobile phonepositioning data, personal travel trajectory data, and patient disease data. How to effectivelyprotect personal privacy and human rights during information acquisition and AI-based processingis an issue worthy of discussion and attention.
In addition to the applications investigated in this paper, AI can also contribute to the battle ofCOVID-19 from the following 10 potential directions.1. Noncontact disease detection. In CXR and CT image detection, the use of noncontact automaticimage acquisition can effectively avoid the risk of infection between radiologists and patients duringthe COVID-19 pandemic. AI can be used for patient posture positioning, standard sectionacquisition of CXR and CT images, and movement of camera equipment.2. Remote video diagnosis. AI and NLP technologies can be used to develop remote video diagnosissystems and chat robot systems and provide COVID-19 disease consultation and preliminarydiagnosis to the public.3. Patient prognosis management. AI technology (such as intelligent image and video analysis) can beused to automatically monitor patient behavior during the follow-up monitoring and prognosticmanagement process, in addition to long-term tracking and management of patients withCOVID-19.4. Biological research. In the field of biological research, AI can be used to discover protein structuresand features of virus through accurate analysis of biomedical information, such as large-scaleprotein structures, gene sequences, and viral trajectories.5. Drug and vaccine development. AI can not only be used to discover potential drugs and vaccinesbut also to simulate the interaction between drugs and proteins and between vaccines andreceptors, thereby predicting the potential responses to the drugs and vaccines of patients withCOVID-19 with different constitutions.6. Identification and filtering of fake news. AI can be used to reduce and eliminate fake news andnoise data on online social media platforms to provide reliable, correct, and scientific informationabout the COVID-19 pandemic.7. Impact simulation and evaluation. Various simulation models can use AI to analyze the impact ofdifferent social control strategies on disease transmission. Then, they can be used to explore moreeffective and scientific approaches of disease prevention and social control.8. Patient contact tracking. By constructing social relationship networks and knowledge graphs, AIcan identify and track the trajectories of people in close contact with patients with COVID-19,thereby accurately predicting and controlling the potential spread of the disease. 22/36. Intelligent robots. Intelligent robots are expected to be used in applications such as disinfectionand cleaning in public places, product distribution, and patient care.10. Intelligent Internet of Things. AI is expected to be combined with the Internet of Things to deployin customs, airports, railway stations, bus stations, and business centers. In this case, we canquickly identify suspicious COVID-19 virus and patients through intelligent monitoring of theenvironment and personnel.
In this survey, we investigated the main scope and contributions of AI in combating COVID-19.Compared with the pandemic of SARS-CoV in 2003 and MERS-CoV in 2012, AI technologies have beensuccessfully applied in almost every corner of the COVID-19 battle. The application of AI in COVID-19research can be summarized in four aspects, such as disease detection and diagnosis, virology research,drug and vaccine development, and epidemic and transmission prediction. Among them, medical imageanalysis, drug discovery, and epidemic prediction are the main battlefields of AI in the fight againstCOVID-19. We also summarized the currently available data and resources for COVID-19 research basedon AI, including medical imaging data, biological data, and informatics resources. Finally, we highlightedthe main challenges and potential directions in this field. This survey provided medical and AIresearchers with a comprehensive view of the existing and potential contributions of AI in combatingCOVID-19, with the goal of inspiring them to continue to maximize the advantages of AI and big datato fight against this pandemic.
Acknowledgment
This work is partially funded by the National Key R&D Program of China (Grant No.2018YFB1003401), the National Outstanding Youth Science Program of National Natural ScienceFoundation of China (Grant No. 61625202), and the International Postdoctoral Exchange FellowshipProgram (Grant No. 20180024).
References
1. T. Abel and D. McQueen. The covid-19 pandemic calls for spatial distancing and social closeness:not for social distancing.
International Journal of Public Health , 65(3):231–231, 2020.2. P. Afshar, S. Heidarian, F. Naderkhani, A. Oikonomou, K. N. Plataniotis, and A. Mohammadi.Covid-caps: a capsule network-based framework for identification of covid-19 cases from x-rayimages. arXiv preprint arXiv:2004.02696 , 2020.3. M. Al, A. Ewees, and H. Fan. Optimization method for forecasting confirmed cases of covid-19 inchina.
Journal of Clinical Medicine , 9(3):674, 2020.4. Z. Allam, G. Dey, and D. Jones. Artificial intelligence (ai) provided early detection of thecoronavirus (covid-19) in china and will influence future urban health policy internationally. AI ,1(2):156–165, 2020.5. A. Alqudah and S. Qazan. Augmented covid-19 x-ray images dataset. Mendeley Data, May 2020. https://data.mendeley.com/datasets/2fxz4px6d8/4 .6. S. Altschul, W. Gish, W. Miller, E. Myers, and D. Lipman. Basic local alignment search tool. Journal of molecular biology , 215(3):403–410, 1990.7. K. Andersen, A. Rambaut, W. Lipkin, E. Holmes, and R. Garry. The proximal origin ofsars-cov-2.
Nature Medicine , 26(4):450–452, 2020. 23/36. M. Andreatta and M. Nielsen. Gapped sequence alignment using artificial neural networks:application to the mhc class i system.
Bioinformatics , 32(4):511–517, 2016.9. I. Apostolopoulos and T. Mpesiana. Covid-19: automatic detection from x-ray images utilizingtransfer learning with convolutional neural networks.
Physical and Engineering Sciences inMedicine , page 1, 2020.10. Y. Arabi, S. Murthy, and S. Webb. Covid-19: a novel coronavirus and a novel challenge forcritical care.
Intensive Care Medicine , pages 1–4, 2020.11. S. Armato, G. McLennan, L. Bidaut, M. McNitt, and C. Meyer. The lung image databaseconsortium (lidc) and image database resource initiative (idri): a completed reference database oflung nodules on ct scans.
Medical Physics , 38(2):915–931, 2011.12. K. Arnold, L. Bordoli, and J. Kopp. The swiss-model workspace: a web-based environment forprotein structure homology modelling.
Bioinformatics , 22(2):195–201, 2006.13. M. Awad and R. Khanna. Support vector regression. In
Efficient Learning Machines , pages67–80. 2015.14. S. Ayyoubzadeh and H. Zahedi. Predicting covid-19 incidence using google trends and datamining techniques: a pilot study in iran.
JMIR Public Health and Surveillance , 2020.15. Bachir. Covid-19 x-rays. Website, May 2020. .16. Baidu. Real-time covid-19 data. Website, May 2020. https://voice.baidu.com/act/newpneumonia .17. A. Bairoch, R. Apweiler, C. Wu, and W. Barker. The universal protein resource (uniprot).
Nucleic Acids Research , 33(suppl 1):D154–D159, 2005.18. B. Beck, B. Shin, Y. Choi, and S. Park. Predicting commercially available antiviral drugs thatmay act on the novel coronavirus (sars-cov-2) through a drug-target interaction deep learningmodel.
Computational and Structural Biotechnology Journal , 2020.19. H. Berman, K. Henrick, and H. Nakamura. Announcing the worldwide protein data bank.
NatureStructural and Molecular Biology , 10(12):980–980, 2003.20. M. Bing. Covid-19 tracker). Website, May 2020. https://bing.com/covid .21. BlueDot. An ai epidemiologist sent the first warnings of the wuhan virus. Website, May 2020. https://bluedot.global .22. L. Breiman. Random forests.
Machine Learning , 45(1):5–32, 2001.23. N. Bung, S. Krishnan, G. Bulusu, and A. Roy. De novo design of new chemical entities (nces) forsars-cov-2 using artificial intelligence. 2020.24. A. C. Covid-19 chest x-ray dataset initiative. Website, May 2020. https://github.com/agchung/Figure1-COVID-chestxray-dataset .25. D. Caccavo. Chinese and italian covid-19 outbreaks can be correctly described by a modified sirdmodel. medRxiv , 2020.26. I. Castiglioni, D. Ippolito, and M. Interlenghi. Artificial intelligence applied on chest x-ray canaid in the diagnosis of covid-19 infection: a first experience from lombardy, italy. medRxiv , 2020.24/367. J. Chan, S. Yuan, K. Kok, and K. To. A familial cluster of pneumonia associated with the 2019novel coronavirus indicating person-to-person transmission: a study of a family cluster.
TheLancet , 395(10223):514–523, 2020.28. S. Chang, N. Harding, C. Zachreson, and O. Cliff. Modelling transmission and control of thecovid-19 pandemic in australia. arXiv preprint arXiv:2003.10218 , 2020.29. T. Chebet, Y. Li, N. Sam, and Y. Liu. A comparative study of fine-tuning deep learning modelsfor plant disease identification.
Computers and Electronics in Agriculture , 161:272–279, 2019.30. H. Chen, J. Guo, C. Wang, F. Luo, and X. Yu. Clinical characteristics and intrauterine verticaltransmission potential of covid-19 infection in nine pregnant women: a retrospective review ofmedical records.
The Lancet , 395(10226):809–815, 2020.31. J. Chen, K. Li, Q. Deng, K. Li, and Y. Philip. Distributed deep learning model for intelligentvideo surveillance systems with edge computing.
IEEE Transactions on Industrial Informatics ,2019.32. J. Chen, L. Wu, and J. Zhang. Deep learning-based model for detecting 2019 novel coronaviruspneumonia on high-resolution computed tomography: a prospective study. medRxiv , 2020.33. L. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. Yuille. Deeplab: Semantic imagesegmentation with deep convolutional nets, atrous convolution, and fully connected crfs.
IEEETransactions on Pattern Analysis and Machine Intelligence , 40(4):834–848, 2017.34. Q. Chen, A. Allot, and Z. Lu. Keep up with the latest coronavirus research.
Nature ,579(7798):193–193, 2020.35. S. Chen, J. Yang, W. Yang, and C. Wang. Covid-19 control in china during mass populationmovements at new year.
The Lancet , 395(10226):764–766, 2020.36. S. Chen, Z. Zhang, J. Yang, J. Wang, and X. Zhai. Fangcang shelter hospitals: a novel conceptfor responding to public health emergencies.
The Lancet , 2020.37. T. Chen and C. Guestrin. Xgboost: a scalable tree boosting system. In
ACM KDD , pages785–794, 2016.38. W. Chen, U. Strych, P. Hotez, and M. Bottazzi. The sars-cov-2 vaccine pipeline: an overview.
Current Tropical Medicine Reports , pages 1–4, 2020.39. F. Chollet. Xception: deep learning with depthwise separable convolutions. In
IEEE CVPR ,pages 1251–1258, 2017.40. A. Chun. In a time of coronavirus, china’s investment in ai is paying off in a big way.
SouthChina Morning Post , 18, 2020.41. J. Cohen, P. Morrison, and L. Dao. Covid-19 image data collection. arXiv preprintarXiv:2003.11597 , 2020.42. Coronacases. Ct images of confirmed covid-19 cases. Mendeley Data, May 2020. https://coronacases.org .43. S. D. Novel corona virus 2019 dataset. Website, May 2020. .44. Y. Decastro, F. Gamboa, D. Henrion, and R. Hess. Approximate optimal designs for multivariatepolynomial regression.
The Annals of Statistics , 47(1):127–155, 2019. 25/365. J. Degen, C. Wegscheid, A. Zaliani, and M. Rarey. On the art of compiling andusing’drug-like’chemical fragment spaces.
ChemMedChem: Chemistry Enabling Drug Discovery ,3(10):1503–1507, 2008.46. M. Demirci and A. Adan. Computational analysis of microrna-mediated interactions in sars-cov-2infection. bioRxiv , 2020.47. J. Deng, W. Dong, R. Socher, and L. Li. Imagenet: a large-scale hierarchical image database. In
IEEE CVPR , pages 248–255, 2009.48. A. Depeursinge, A. Vargas, and A. Platon. Building a reference multimedia database forinterstitial lung diseases.
Computerized Medical Imaging and Graphics , 36(3):227–238, 2012.49. E. Dong, H. Du, and L. Gardner. An interactive web-based dashboard to track covid-19 in realtime.
The Lancet Infectious Diseases , 2020.50. Eurorad. Images of covid-19 cases. Mendeley Data, May 2020. .51. S. Fong, G. Li, N. Dey, and R. Crespo. Composite monte carlo decision making under highuncertainty of novel coronavirus epidemic using hybridized deep learning and fuzzy ruleinduction.
Applied Soft Computing , page 106282, 2020.52. S. Fong, G. Li, N. Dey, and R. Crespo. Finding an accurate early forecasting model from smalldataset: a case of 2019-ncov novel coronavirus outbreak. arXiv preprint arXiv:2003.10776 , 2020.53. N. C. for Biotechnology Information (NCBI). Genome sequencing data of sars-cov-2. Website,May 2020. .54. C. for Disease Control and Prevention. Weekly influenza confirmed cases. Website, May 2020. .55. E. C. for Disease Prevention and Control. Geographic distribution covid-19 cases worldwide.Website, May 2020. https://ecdc.europa.eu/en/publications-data .56. A. Gaulton, L. Bellis, A. Bento, J. Chambers, and M. Davies. Chembl: a large-scale bioactivitydatabase for drug discovery.
Nucleic Acids Research , 40(D1):D1100–D1107, 2012.57. R. Ge, Y. Qi, Y. Yan, M. Tian, and Q. Gu. The role of close contacts tracking management incovid-19 prevention: a cluster investigation in jiaxing, china.
The Journal of Infection , 2020.58. G. H. D. D. I. (GHDDI). Targeting covid-19: Ghddi info sharing portal. Website, May 2020. https://ghddi-ailab.github.io/Targeting2019-nCoV .59. I. Ghinai, T. McPherson, J. Hunter, and H. Kirking. First known person-to-person transmissionof severe acute respiratory syndrome coronavirus 2 (sars-cov-2) in the usa.
The Lancet , 2020.60. M. Gilman, C. Liu, A. Fung, and I. Behera. Structure of the respiratory syncytial viruspolymerase complex.
Cell , 179(1):193–204, 2019.61. GISAID. Gisaid: global initiative on sharing all influenza data. Website, May 2020. .62. D. Giuliani, M. Dickson, G. Espa, and F. Santi. Modelling and predicting the spatio-temporalspread of coronavirus disease 2019 (covid-19) in italy.
SSRN , 2020.63. R. G´omez, J. Wei, D. Duvenaud, and J. Hern´andez. Automatic chemical design using adata-driven continuous representation of molecules.
ACS Central Science , 4(2):268–276, 2018.64. GOV.UK. Coronavirus (covid-19) cases in the uk. Website, May 2020. https://coronavirus.data.gov.uk/ . 26/365. O. Gozes, M. Frid, and H. Greenspan. Rapid ai development cycle for the coronavirus (covid-19)pandemic: Initial results for automated detection and patient monitoring using deep learning ctimage analysis. arXiv preprint arXiv:2003.05037 , 2020.66. K. Greff, R. Srivastava, K. Jan, and B. Steunebrink. Lstm: a search space odyssey.
IEEETransactions on Neural Networks and Learning Systems , 28(10):2222–2232, 2016.67. A. Hassanien, L. Mahdy, K. Ezzat, and H. Elmousalami. Automatic x-ray covid-19 lung imageclassification system based on multi-level thresholding and support vector machine. medRxiv ,2020.68. H. D. E. (HDE). Covid-19 pandemic. Website, May 2020. https://data.humdata.org/event/COVID-19 .69. K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In
IEEECVPR , pages 770–778, 2016.70. X. He, E. Lau, P. Wu, X. Deng, and J. Wang. Temporal dynamics in viral shedding andtransmissibility of covid-19.
Nature Medicine , pages 1–4, 2020.71. Y. He, Z. Xiang, and H. Mobley. Vaxign: the first web-based vaccine design program for reversevaccinology and applications for vaccine development.
BioMed Research International , 2010,2010.72. J. Hellewell, S. Abbott, A. Gimma, and N. Bosse. Feasibility of controlling covid-19 outbreaks byisolation of cases and contacts.
The Lancet Global Health , 2020.73. E. Hemdan, M. Shouman, and M. Karar. Covidx-net: A framework of deep learning classifiers todiagnose covid-19 in x-ray images. arXiv preprint arXiv:2003.11055 , 2020.74. C. Herst, S. Burkholz, J. Sidney, A. Sette, P. Harris, S. Massey, T. Brasel, E. Cunha, D. S. Rosa,and W. Chao. An effective ctl peptide vaccine for ebola zaire based on survivors’ cd8+ targetingof a particular nucleocapsid protein epitope with potential implications for covid-19 vaccinedesign.
Vaccine , 2020.75. G. Hinton, S. Sabour, and N. Frosst. Matrix capsules with em routing. 2018.76. S. Hochreiter and J. Schmidhuber. Long short-term memory.
Neural Computation ,9(8):1735–1780, 1997.77. M. Hoffmann, H. Kleine, S. Schroeder, and N. Kr¨uger. Sars-cov-2 cell entry depends on ace2 andtmprss2 and is blocked by a clinically proven protease inhibitor.
Cell , 2020.78. M. Hofmarcher, A. Mayr, E. Rumetshofer, and P. Ruch. Large-scale ligand-based virtualscreening for sars-cov-2 inhibitors using deep neural networks.
SSRN , 2020.79. F. Hu, J. Jiang, and P. Yin. Prediction of potential commercially inhibitors against sars-cov-2 bymulti-task deep model. arXiv preprint arXiv:2003.00728 , 2020.80. Z. Hu, Q. Ge, L. Jin, and M. Xiong. Artificial intelligence forecasting of covid-19 in china. arXivpreprint arXiv:2002.07112 , 2020.81. Z. Hu, Q. Ge, S. Li, L. Jin, and M. Xiong. Evaluating the effect of public health intervention onthe global-wide spread trajectory of covid-19. medRxiv , 2020.82. C. Huang, Y. Chen, Y. Ma, and P. Kuo. Multiple-input deep convolutional neural network modelfor covid-19 forecasting inchina. medRxiv , 2020. 27/363. C. Huang, Y. Wang, X. Li, L. Ren, and J. Zhao. Clinical features of patients infected with 2019novel coronavirus in wuhan, china.
The Lancet , 395(10223):497–506, 2020.84. G. Huang, Z. Liu, V. D, and K. Weinberger. Densely connected convolutional networks. In
IEEECVPR , pages 4700–4708, 2017.85. L. Huang, R. Han, T. Ai, P. Yu, and H. Kang. Serialquantitative chest ct assessment of covid-19:deep-learning approach.
Radiology: Cardiothoracic Imaging , 2(2):e200075, 2020.86. G. Ian, P. Jean, M. Mirza, B. Xu, and W. David. Generative adversarial nets. In
Advances inNeural Information Processing Systems , pages 2672–2680, 2014.87. IEDB. Ellipro: an antibody epitope prediction tool. Website, May 2020. http://tools.iedb.org/ellipro .88. G. Inc. Github. Website, May 2020. https://github.com .89. C. A. Integrity. Clarivate analytics integrity. Website, May 2020. https://integrity.clarivate.com/integrity/ .90. M. Iqbal. Active surveillance for covid-19 through artificial intelligence using concept of real-timespeech-recognition mobile application to analyse cough sound. 2020.91. S. Jaeger, S. Fulle, and S. Turk. Mol2vec: unsupervised machine learning approach with chemicalintuition.
Journal of Chemical Information and Modeling , 58(1):27–35, 2018.92. J. Jang. Anfis: adaptive-network-based fuzzy inference system.
IEEE Transactions on Systems,Man, and Cybernetics , 23(3):665–685, 1993.93. H. Jeffrey. Chaos game representation of gene structure.
Nucleic Acids Research ,18(8):2163–2170, 1990.94. C. Jin, W. Chen, Y. Cao, Z. Xu, X. Zhang, and L. Deng. Development and evaluation of an aisystem for covid-19 diagnosis. medRxiv , 2020.95. Y. Jin, H. Yang, W. Ji, W. Wu, and S. Chen. Virology, epidemiology, pathogenesis, and controlof covid-19.
Viruses , 12(4):372, 2020.96. J. Jumper, K. Tunyasuvunakool, P. Kohli, and D. Hassabis. Computational predictions ofprotein structures associated with covid-19. Website, May 2020. https://deepmind.com .97. V. Jurtz, S. Paul, M. Andreatta, P. Marcatili, B. Peters, and M. Nielsen. Netmhcpan-4.0:improved peptide-mhc class i interaction predictions integrating eluted ligand and peptidebinding affinity data.
The Journal of Immunology , 199(9):3360–3368, 2017.98. O. Kadioglu, M. Saeed, H. Johannes, and T. Efferth. Identification of novel compounds againstthree targets of sars cov-2 coronavirus by combined virtual screening and supervised machinelearning.
Bulletin of the World Health Organization , 2020.99. N. Kandel, S. Chungong, A. Omaar, and J. Xing. Health security capacities in the context ofcovid-19 outbreak: an analysis of international health regulations annual report data from 182countries.
The Lancet , 2020.100. M. Keeling, T. Hollingsworth, and J. Read. The efficacy of contact tracing for the containment ofthe 2019 novel coronavirus(covid-19). medRxiv , 2020.101. P. Keeling. Modeling infectious diseases in humans and animals.
Anonymous Princeton:Princeton University Press , 366:362, 2008. 28/3602. D. Kermany, M. Goldbaum, and W. Cai. Identifying medical diagnoses and treatable diseases byimage-based deep learning.
Cell , 172(5):1122–1131, 2018.103. J. Kim, M. Kang, E. Park, and D. Chung. A simple and multiplex loop-mediated isothermalamplification assay for rapid detectionof sars-cov.
BioChip Journal , 13(4):341–351, 2019.104. S. Kim, P. Thiessen, E. Bolton, and J. Chen. Pubchem substance and compound databases.
Nucleic Acids Research , 44(D1):D1202–D1213, 2016.105. S. Kissler, C. Tedijanto, M. Lipsitch, and Y. Grad. Social distancing strategies for curbing thecovid-19 epidemic. medRxiv , 2020.106. W. Kong, Y. Li, M. Peng, and D. Kong. Sars-cov-2 detection in patients with influenza-likeillness.
Nature Microbiology , pages 1–4, 2020.107. J. Koo, A. Cook, M. Park, Y. Sun, and H. Sun. Interventions to mitigate early spread ofsars-cov-2 in singapore: a modelling study.
The Lancet Infectious Diseases , 2020.108. A. Kozomara, M. Birgaoanu, and S. Griffiths. mirbase: from microrna sequences to function.
Nucleic Acids Research , 47(D1):D155–D162, 2019.109. A. Krizhevsky, I. Sutskever, and G. Hinton. Imagenet classification with deep convolutionalneural networks. In
Advances in Neural Information Processing Systems , pages 1097–1105, 2012.110. S. Lai, I. Bogoch, and N. Ruktanonchai. Assessing spread risk of wuhan novel coronavirus withinand beyond china, january-april 2020: a travel network-based modelling study. 2020.111. V. Lampos, S. Moura, E. Yom, and I. Cox. Tracking covid-19 using online search. arXiv preprintarXiv:2003.08086 , 2020.112. Larxel. Covid-19 x-rays. Website, May 2020. .113. G. Li and E. De. Therapeutic options for the 2019 novel coronavirus (2019-ncov). 2020.114. L. Li, L. Qin, Z. Xu, and Y. Yin. Artificial intelligence distinguishes covid-19 from communityacquired pneumonia on chest ct.
Radiology , page 200905, 2020.115. L. Li, Q. Zhang, X. Wang, J. Zhang, and T. Wang. Characterizing the propagation of situationalinformation in social media during covid-19 epidemic: a case study on weibo.
IEEE Transactionson Computational Social Systems , 2020.116. S. Li, K. Song, B. Yang, Y. Gao, and X. Gao. Preliminary assessment of the covid-19 outbreakusing 3-staged model e-ishr.
Journal of Shanghai Jiaotong University (Science) , 25:157–164, 2020.117. Y. Li, Z. Zhu, Y. Zhou, Y. Xia, and W. Shen. Volumetric medical image segmentation: a 3d deepcoarse-to-fine framework and its adversarial examples. In
Deep Learning and ConvolutionalNeural Networks for Medical Imaging and Clinical Informatics , pages 69–91. Springer, 2019.118. J. Liu, R. Cao, M. Xu, X. Wang, and H. Zhang. Hydroxychloroquine, a less toxic derivative ofchloroquine, is effective in inhibiting sars-cov-2 infection in vitro.
Cell Discovery , 6(1):1–4, 2020.119. P. Liu, P. Beeler, and R. Chakrabarty. Covid-19progression timeline and effectiveness ofresponse-to-spread interventions across the united states. medRxiv , 2020.120. T. Liu, Y. Lin, X. Wen, R. N. Jorissen, and M. Gilson. Bindingdb: a web-accessible database ofexperimentally determined protein-ligand binding affinities.
Nucleic Acids Research ,35(suppl 1):D198–D201, 2007. 29/3621. M. Loey, F. Smarandache, and N. Khalifa. Within the lack of covid-19 benchmark dataset: anovel gan with deep transfer learning for corona-virus detection in chest x-ray images. 2020.122. R. Lu, X. Wu, Z. Wan, Y. Li, and L. Zuo. Development of anovel reverse transcriptionloop-mediated isothermal amplification method for rapid detection ofsars-cov-2.
VirologicaSinica , page 1, 2020.123. R. Lu, X. Zhao, J. Li, P. Niu, B. Yang, and H. Wu. Genomic characterisation and epidemiologyof 2019 novel coronavirus: implications for virus origins and receptor binding.
The Lancet ,395(10224):565–574, 2020.124. P. M. Chest x-ray images (pneumonia). Website, May 2020. .125. L. Maaten and G. Hinton. Visualizing data using t-sne.
Journal of Machine Learning Research ,9(Nov):2579–2605, 2008.126. A. magazine. International air travel association(iata). Website, May 2020. .127. H. Maghdid, A. Asaad, K. Ghafoor, A. Sadiq, and M. Khan. Diagnosing covid-19 pneumoniafrom x-ray and ct images using deep learning and transfer learning algorithms. arXiv preprintarXiv:2004.00038 , 2020.128. H. Maghdid, K. Ghafoor, A. Sadiq, K. Curran, and K. Rabie. A novel ai-enabled framework todiagnose coronavirus covid 19 using smartphone embedded sensors: Design study. arXiv preprintarXiv:2003.07434 , 2020.129. M. Marini, C. Brunner, N. Chokani, and R. Abhari. Enhancing response preparedness toinfluenza epidemics: agent-based study of 2050 influenza season in switzerland.
SimulationModelling Practice and Theory , page 102091, 2020.130. M. Marini, N. Chokani, and R. Abhari. Covid-19 epidemic in switzerland: growth prediction andcontainment strategy using artificial intelligence and big data. medRxiv , 2020.131. MediaCloud. Mediacloud. Website, May 2020. .132. Metabiota. How ai is battling the coronavirus outbreak. Website, May 2020. .133. H. Metsky, C. Freije, and T. Kosoko. Crispr-based surveillance for covid-19 usingkosoko-thoroddsen, tinna-solveig f-comprehensive machine learning design. bioRxiv , 2020.134. H. Mi, A. Muruganujan, and P. Thomas. Panther in 2013: modeling the evolution of genefunction, and other gene attributes, in the context of phylogenetic trees.
Nucleic Acids Research ,41(D1):D377–D386, 2012.135. T. Mihara, Y. Nishimura, Y. Shimizu, H. Nishiyama, G. Yoshikawa, H. Uehara, P. Hingamp,S. Goto, and H. Ogata. Linking virus genomes with host taxonomy.
Viruses , 8(3):66, 2016.136. M. Moskal, W. Beker, R. Roszak, and E. Gajewska. Suggestions for second-pass anti-covid-19drugs based on the artificial intelligence measures of molecular similarity, shape andpharmacophore distribution. 2020.137. A. Narin, C. Kaya, and Z. Pamuk. Automatic detection of coronavirus disease (covid-19) usingx-ray images and deep convolutional neural networks. arXiv preprint arXiv:2003.10849 , 2020.138. R. S. of North America. Rsna pneumonia detection challenge. Website, May 2020. . 30/3639. N. H. C. of the People’s Republic of China. Real-time covid-19 report. Website, May 2020. .140. B. S. of Thoracic Imaging. Covid-19 bsti imaging database. Website, May 2020. .141. S. Oh, W. Pedrycz, and B. Park. Polynomial neural networks architecture: analysis and design.
Computers and Electrical Engineering , 29(6):703–725, 2003.142. E. Ong, H. Wang, M. Wong, and M. Seetharaman. Vaxign-ml: supervised machine learningreverse vaccinology model for improved prediction of bacterial protective antigens.
Bioinformatics , 2020.143. E. Ong, M. Wong, A. Huffman, and Y. He. Covid-19 coronavirus vaccine design using reversevaccinology and machine learning.
BioRxiv , 2020.144. I. S. . ore. The coronavirus datasets in italy. Website, May 2020. https://lab24.ilsole24ore.com/coronavirus .145. J. Ortega, M. Serrano, F. Pujol, and H. Rangel. Role of changes in sars-cov-2 spike protein in theinteraction with the human ace2 receptor: nn in silico analysis.
EXCLI Journal , 19:410, 2020.146. X. Ou, Y. Liu, X. Lei, P. Li, D. Mi, L. Ren, and L. Guo. Characterization of spike glycoproteinof sars-cov-2 on virus entry and its immune cross-reactivity with sars-cov.
NatureCommunications , 11(1):1–12, 2020.147. X. Pan, X. Da, W. Zhou, and L. Wang. Identification of a potential mechanism of acute kidneyinjury during the covid-19 outbreak: a study based on single-cell transcriptome analysis.
Intensive Care Medicine , page 1, 2020.148. T. paper. The paper news network. Website, May 2020. .149. R. P. D. B. (PDB). Covid-19/sars-cov-2 resources. Website, May 2020. .150. J. Phua, L. Weng, L. Ling, M. Egi, and C. Lim. Intensive care management of coronavirus disease2019 (covid-19): challenges and recommendations.
The Lancet Respiratory Medicine , 2020.151. B. Pierce, K. Wiehe, H. Hwang, and B. Kim. Zdock server: interactive docking prediction ofprotein-protein complexes and symmetric multimers.
Bioinformatics , 30(12):1771–1773, 2014.152. M. Pourhomayoun and M. Shakibi. Predicting mortality risk in patients with covid-19 usingartificial intelligence to help medical decision-making. medRxiv , 2020.153. M. Prachar, S. Justesen, D. Steen, S. Thorgrimsen, E. Jurgons, O. Winther, and F. Bagger.Covid-19 vaccine candidates: Prediction and validation of 174 sars-cov-2 epitopes. bioRxiv , 2020.154. K. Preuer, G. Klambauer, F. Rippmann, and S. Hochreiter. Interpretable deep learning in drugdiscovery. In
Explainable AI , pages 331–345. Springer, 2019.155. N. Punn, S. Sonbhadra, and S. Agarwal. Covid-19 epidemic analysis using machine learning anddeep learning algorithms. medRxiv , 2020.156. X. Qi, Z. Jiang, Q. Yu, C. Shao, and H. Zhang. Machine learning-based ct radiomics model forpredicting hospital stay in patients with pneumonia associated with sars-cov-2 infection: amulticenter study. medRxiv , 2020.157. X. Qian, R. Ren, Y. Wang, Y. Guo, and J. Fang. Fighting against the common enemy ofcovid-19: a practice of building a community with a shared future for mankind.
InfectiousDiseases of Poverty , 9(1):1–6, 2020. 31/3658. R. Qiao, N. Tran, B. Shan, A. Ghodsi, and M. Li. Personalized workflow to identify optimal t-cellepitopes for peptide-based vaccines against covid-19. arXiv preprint arXiv:2003.10650 , 2020.159. A. R. Detecting covid-19 in x-ray images with keras, tensorflow, and deep learning. Website,May 2020. .160. T. R. Novel corona virus 2019 dataset. Website, May 2020. .161. Radiopaedia. Images of covid-19 cases. Mendeley Data, May 2020. https://radiopaedia.org .162. M. Rahman, M. Hoque, M. Islam, S. Akter, A. Rubayet, M. Siddique, O. Saha, M. Rahaman,M. Sultana, and M. Hossain. Epitope-based chimeric peptide vaccine design against s, m and eproteins of sars-cov-2 etiologic agent of global pandemic covid-19: an in silico approach.
BioRxiv ,2020.163. S. Rahmatizadeh, S. Valizadeh, and A. Dabbagh. The role of artificial intelligence in managementof critical covid-19 patients.
Journal of Cellular and Molecular Anesthesia , 5(1):16–22, 2020.164. G. Randhawa, M. Soltysiak, H. El, and C. de Souza. Machine learning using intrinsic genomicsignatures for rapid classification of novel pathogens: Covid-19 case study. bioRxiv , 2020.165. A. Rao and J. Vazquez. Identification of covid-19 can be quicker through artificial intelligenceframework using a mobile phone-based survey in the populations when cities/towns are underquarantine.
Infection Control and Hospital Epidemiology , pages 1–18, 2020.166. N. Rapin, O. Lund, M. Bernaschi, and F. Castiglione. Computational immunology meetsbioinformatics: the use of prediction tools for molecular binding in the simulation of the immunesystem.
PLoS One , 5(4), 2010.167. A. Rivas. Drones and artificial intelligence to enforce social isolation during covid-19 outbreak.
Medium: Towards Data Science , 26, 2020.168. R. Rizk and A. Hassanien. Covid-19 forecasting based on an improved interior search algorithmand multi-layer feed forward neural network. arXiv preprint arXiv:2004.05960 , 2020.169. R. Sameni. Mathematical modeling of epidemic diseases; a case study of the covid-19 coronavirus. arXiv preprint arXiv:2003.11371 , 2020.170. M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L. Chen. Mobilenetv2: inverted residualsand linear bottlenecks. In
IEEE CVPR , pages 4510–4520, 2018.171. K. Santosh. Ai-driven tools for coronavirus outbreak: need of active learning and cross-populationtrain/test models on multitudinal/multimodal data.
Journal of Medical Systems , 44(5):1–5, 2020.172. B. Sarkar, M. Ullah, F. Johora, and M. Taniya. The essential facts of wuhan novel coronavirusoutbreak in china and epitope-based vaccine designing against 2019-ncov.
BioRxiv , 2020.173. J. Sarkar and P. Chakrabarti. A machine learning model reveals older age and delayedhospitalization as predictors of mortality in patients with covid-19. medRxiv , 2020.174. B. Schuller, D. Schuller, K. Qian, J. Liu, H. Zheng, and X. Li. Covid-19 and computer audition:An overview on what speech and sound analysis could contribute in the sars-cov-2 corona crisis. arXiv preprint arXiv:2003.11117 , 2020.175. M. Segmentation. Covid-19 ct segmentation dataset. Website, May 2020. http://medicalsegmentation.com/covid19 . 32/3676. A. Senior, R. Evans, J. Jumper, and J. Kirkpatrick. Improved protein structure prediction usingpotentials from deep learning.
Nature , pages 1–5, 2020.177. A. Senior, R. Evans, J. Jumper, J. Kirkpatrick, and L. Sifre. Protein structure prediction usingmultiple deep neural networks casp13.
Proteins: Structure, Function, and Bioinformatics ,87(12):1141–1148, 2019.178. P. Sethy and S. Behera. Detection of coronavirus disease (covid-19) based on deep features. , 2020030300:2020, 2020.179. F. Shan, Y. Gao, J. Wang, W. Shi, and N. Shi. Lung infection quantification of covid-19 in ctimages with deep learning. arXiv preprint arXiv:2003.04655 , 2020.180. F. Shi, L. Xia, F. Shan, and D. Wu. Large-scale screening of covid-19 from community acquiredpneumonia using infection size-aware classification. arXiv preprint arXiv:2003.09860 , 2020.181. W. Shi, X. Peng, T. Liu, and Z. Cheng. Deep learning-based quantitative computed tomographymodel in predicting the severity of covid-19: a retrospective study in 196 patients.
SSRN , 2020.182. B. Shin, S. Park, K. Kang, and J. Ho. Self-attention based molecule representation for predictingdrug-target interaction. arXiv preprint arXiv:1908.06760 , 2019.183. K. Simonyan and A. Zisserman. Very deep convolutional networks for large-scale imagerecognition. arXiv preprint arXiv:1409.1556 , 2014.184. SIRM. Covid-19 database. Website, May 2020. https://sirm.org/category/senza-categoria/COVID-19 .185. M. Siwiak, P. Szczesny, and M. Siwiak. From a single host to global spread. the global mobilitybased modelling of the covid-19 pandemic implies higher infection and lower detection rates thancurrent estimates.
The Global Mobility Based Modelling , 2020.186. M. Skalic, J. Jim´enez, D. Sabbadin, and G. De. Shape-based generative modeling for de novodrug design.
Journal of Chemical Information and Modeling , 59(3):1205–1214, 2019.187. Y. Song, S. Zheng, L. Li, X. Zhang, and X. Zhang. Deep learning enables accurate diagnosis ofnovel coronavirus (covid-19) with ct images. medRxiv , 2020.188. M. Stepniewska, P. Zielenkiewicz, and P. Siedlecki. Development and evaluation of a deeplearning model for protein-ligand binding affinity prediction.
Bioinformatics , 34(21):3666–3674,2018.189. C. Szegedy, S. Ioffe, V. Vanhoucke, and A. Alemi. Inception-v4, inception-resnet and the impactof residual connections on learning. In
AAAI Conference on Artificial Intelligence , 2017.190. C. Szegedy, W. Liu, and P. Sermanet. Going deeper with convolutions. In
IEEE CVPR , pages1–9, 2015.191. C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, Z. Wojna, and Z. Wojna. Rethinking the inceptionarchitecture for computer vision. In
IEEE CVPR , pages 2818–2826, 2016.192. B. Tang, F. He, D. Liu, M. Fang, Z. Wu, and D. Xu. Ai-aided design of novel targeted covalentinhibitors against sars-cov-2. bioRxiv , 2020.193. Z. Tang, W. Zhao, X. Xie, and Z. Zhong. Severity assessment of coronavirus disease 2019(covid-19) using quantitative features from chest ct images. arXiv preprint arXiv:2003.11988 ,2020. 33/3694. Z. Tanoli, Z. Alam, and V. Markus. Drug target commons 2.0: a community platform forsystematic analysis of drug-target interaction profiles.
Database , 2018, 2018.195. P. Teles. Predicting the evolution of sars-covid-2 in portugal using an adapted sir modelpreviously used in south korea for the mers outbreak. arXiv preprint arXiv:2003.10047 , 2020.196. I. Thevarajan, T. Nguyen, M. Koutsakos, J. Druce, L. Caly, C. Sandt, X. Jia, S. Nicholson,M. Catton, and B. Cowie. Breadth of concomitant immune responses prior to patient recovery: acase report of non-severe covid-19.
Nature Medicine , 26(4):453–455, 2020.197. H. Tian, Y. Liu, Y. Li, and C. Wu. An investigation of transmission control measures during thefirst 50 days of the covid-19 epidemic in china.
Science , 2020.198. K. To, O. Tsang, W. Leung, and A. Tam. Temporal profiles of viral load in posteriororopharyngeal saliva samples and serum antibody responses during infection by sars-cov-2: anobservational cohort study.
The Lancet Infectious Diseases , 2020.199. A. Toda. Susceptible-infected-recovered (sir) dynamics of covid-19 and economic impact. arXivpreprint arXiv:2003.11221 , 2020.200. F. Touret and X. Lamballerie. Of chloroquine and covid-19.
Antiviral research , page 104762,2020.201. G. Trends. Coronavirus search trends. Website, May 2020. https://trends.google.com/trends/story/US_cu_4Rjdh3ABAABMHM_en .202. O. Trott and A. Olson. Autodock vina: improving the speed and accuracy of docking with a newscoring function, efficient optimization, and multithreading.
Journal of Computational Chemistry ,31(2):455–461, 2010.203. R. Viner, S. Russell, H. Croker, J. Packer, J. Ward, C. Stansfield, O. Mytton, and C. Bonell.School closure and management practices during coronavirus outbreaks including covid-19: arapid systematic review.
The Lancet Child and Adolescent Health , 2020.204. R. Vita, L. Zarebski, J. Greenbaum, and H. Emami. The immune epitope database 2.0.
NucleicAcids Research , 38(suppl 1):D854–D862, 2010.205. I. Wallach, M. Dzamba, and A. Heifets. Atomnet: a deep convolutional neural network forbioactivity prediction in structure-based drug discovery. arXiv preprint arXiv:1510.02855 , 2015.206. A. Walls, X. Xiong, and Y. Park. Unexpected receptor functional mimicry elucidates activationof coronavirus fusion.
Cell , 176(5):1026–1039, 2019.207. A. C. Walls, Y. Park, M. Tortorici, and A. Wall. Structure, function, and antigenicity of thesars-cov-2 spike glycoprotein.
Cell , 2020.208. H. Wang, Y. Zhang, S. Lu, and S. Wang. Tracking and forecasting milepost moments of theepidemic in the early-outbreak: framework and applications to the covid-19. medRxiv , 2020.209. L. Wang and A. Wong. Covid-net: A tailored deep convolutional neural network design fordetection of covid-19 cases from chest radiography images. arXiv preprint arXiv:2003.09871 ,2020.210. M. Wang, R. Cao, L. Zhang, and X. Yang. Remdesivir and chloroquine effectively inhibit therecently emerged novel coronavirus (2019-ncov) in vitro.
Cell Research , 30(3):269–271, 2020.211. N. Wang, O. Rosen, and L. Wang. Structural definition of a neutralization-sensitive epitope onthemers-cov s1-ntd.
Cell Reports , 28(13):3395–3405, 2019. 34/3612. S. Wang, B. Kang, J. Ma, and X. Zeng. A deep learning algorithm using ct images to screen forcorona virus disease (covid-19). medRxiv , 2020.213. X. Wang, Y. Peng, L. Lu, and R. Summers. Hospital-scale chest x-ray database and benchmarkson weakly-supervised classification and localization of common thorax diseases. In
IEEE CVPR ,2017.214. Y. Wang, M. Hu, Q. Li, and X. Zhang. Abnormal respiratory patterns classifier may contributeto large-scale screening of people infected with covid-19 in an accurate and unobtrusive manner. arXiv preprint arXiv:2002.05534 , 2020.215. D. Ward, M. Higgins, J. Phelan, M. Hibberd, S. Campino, and T. Clark. An integrated in silicoimmuno-genetic analytical platform provides insights into covid-19 serological and vaccine targets. bioRxiv , 2020.216. WHO. Novel coronavirus 2019 (covid-19). Website, May 2020. .217. R. W¨olfel, V. Corman, W. Guggemos, and M. Seilmaier. Virological assessment of hospitalizedpatients with covid-2019.
Nature , pages 1–10, 2020.218. Worldometer. Covid-19 coronavirus pandemic. Website, May 2020. .219. Worldpop. The statistics datasets on holidays and air travel. Website, May 2020. .220. F. Wu, S. Zhao, B. Yu, Y. Chen, and W. Wang. A new coronavirus associated with humanrespiratory disease in china.
Nature , 579(7798):265–269, 2020.221. J. Wu, K. Leung, M. Bushman, N. Kishore, R. Niehus, and P. Salazar. Estimating clinicalseverity of covid-19 from the transmission dynamics in wuhan, china.
Nature Medicine , pages1–5, 2020.222. J. Wu, K. Leung, and G. Leung. Nowcasting and forecasting the potential domestic andinternational spread of the 2019-ncov outbreak originating in wuhan, china: a modelling study.
The Lancet , 395(10225):689–697, 2020.223. J. Wu, P. Zhang, L. Zhang, W. Meng, and J. Li. Rapid and accurate identification of covid-19infection through machine learning based on clinical available blood test results. medRxiv , 2020.224. S. Xia, M. Liu, C. Wang, W. Xu, and Q. Lan. Inhibition of sars-cov-2 (previously 2019-ncov)infection by a highly potent pan-coronavirus fusion inhibitor targeting its spike protein thatharbors a high capacity to mediate membrane fusion.
Cell Research , pages 1–13, 2020.225. B. Xu, B. Gutierrez, S. Mekaru, and K. Sewalk. Epidemiological data from the covid-19outbreak, real-time case information.
Scientific data , 7(1):1–6, 2020.226. X. Xu, X. Jiang, C. Ma, P. Du, X. Li, and S. Lv. Deep learning system to screen coronavirusdisease 2019 pneumonia. arXiv preprint arXiv:2002.09334 , 2020.227. Y. Xu, X. Li, B. Zhu, H. Liang, and C. Fang. Characteristics of pediatric sars-cov-2 infection andpotential evidence for persistent fecal viral shedding.
Nature Medicine , 26(4):502–505, 2020.228. L. Yan, H. Zhang, J. Goncalves, and Y. Xiao. A machine learning-based model for survivalprediction in patients with severe covid-19 infection. medRxiv , 2020. 35/3629. L. Yan, H. Zhang, Y. Xiao, and M. Wang. Prediction of criticality in patients with severecovid-19 infection using three clinical features: a machine learning-based prognostic model withclinical data in wuhan. medRxiv , 2020.230. H. Yang, W. Xie, X. Xue, and K. Yang. Design of wide-spectrum inhibitors targeting coronavirusmain proteases.
PLoS Biology , 3(10), 2005.231. W. Yang, Q. Cao, L. Qin, X. Wang, and Z. Cheng. Clinical characteristics and imagingmanifestations of the 2019 novel coronavirus disease (covid-19): a multi-center study in wenzhoucity, zhejiang, china.
Journal of Infection , 2020.232. X. Yang. Flower pollination algorithm for global optimization. In
International Conference onUnconventional Computing and Natural Computation , pages 240–249. Springer, 2012.233. Z. Yang, Z. Zeng, and K. Wang. Modified seir and ai prediction of the epidemics trend ofcovid-19 under public health interventions.
Journal of Thoracic Disease , 12(3):165, 2020.234. B. Zareie, A. Roshani, M. Mansournia, and M. Rasouli. A model for covid-19 prediction in iranbased on china parameters.
Archives of Iranian Medicine , 23(4):244–248, 2020.235. C. Zavaleta. Covid-19: review indigenous peoples’ data.
Nature , 580(7802):185, 2020.236. P. Zhai, Y. Ding, X. Wu, and J. Long. The epidemiology, diagnosis and treatment of covid-19.
International Journal of Antimicrobial Agents , page 105955, 2020.237. J. Zhang, Y. Xie, Y. Li, C. Shen, and Y. Xia. Covid-19 screening on chest x-ray images usingdeep learning based anomaly detection. arXiv preprint arXiv:2003.12338 , pages 1–10, 2020.238. J. Zhang, H. Zeng, J. Gu, H. Li, L. Zheng, and Q. Zou. Progress and prospects on vaccinedevelopment against sars-cov-2.
Vaccines , 8(2):153, 2020.239. J. Zhao, Y. Zhang, X. He, and P. Xie. Covid-ct-dataset: a ct scan dataset about covid-19. arXivpreprint arXiv:2003.13865 , 2020. https://github.com/UCSD-AI4H/COVID-CT .240. A. Zhavoronkov, V. Aladinskiy, and A. Zhebrak. Potential covid-2019 3c-like protease inhibitorsdesigned using generative deep learning approaches.
Insilico Medicine , 307:E1, 2020.241. C. Zheng, X. Deng, Q. Fu, Q. Zhou, and J. Feng. Deep learning-based detection for covid-19from chest ct using weak label. medRxiv , 2020.242. P. Zhou, X. Yang, X. Wang, and B. Hu. A pneumonia outbreak associated with a newcoronavirus of probable bat origin.
Nature , 579(7798):270–273, 2020.243. Y. Zhou, Y. Hou, J. Shen, and Y. Huang. Network-based drug repurposing for novel coronavirus2019-ncov/sars-cov-2.
Cell Discovery , 6(1):1–18, 2020.244. Z. Zhou, M. Siddiquee, N. Tajbakhsh, and J. Liang. Unet++: a nested u-net architecture formedical image segmentation. In
Deep Learning in Medical Image Analysis and MultimodalLearning for Clinical Decision Support , pages 3–11. Springer, 2018.245. ZINC. Zinc database. Website, May 2020. https://zinc.docking.orghttps://zinc.docking.org