Development of Machine Learning Regression Model for COVID-19 Drug Target Prediction

There is a perennial need to identify novel, effective therapeutic agents to combat rising infections. Recently, prediction of therapeutic targets to decrease the impact of COVID-19 has posed an urgent challenge requiring innovative solutions. Successful identification of novel drug-target combinations may greatly facilitate drug development. To meet this need, we developed a COVID-19 drug target prediction model using machine learning approaches to quickly identify drug candidates for 18 COVID-19 protein targets. Specifically, we analyzed the performance of three prediction models to predict drug-target docking scores, which represents the strength of interactions between ligands and proteins. Docking scores were predicted for 300,457 molecules on 18 different COVID-19 related protein docking targets. Our proposed approach achieved a competitive performance with mathrm{R}-{2}=0.69,MAE=0.285, MSE=0.627. In addition, we identify chemical structures associated with stronger binding affinities across target binding sites. We believe our work could potentially save pharmaceutical companies significant resources, especially during the early stages of drug development. © 2022 IEEE.

Attention-based Automated Chest CT Image Segmentation Method of COVID-19 Lung Infection

According to the World Health Organization, Artificial Intelligence (AI) technology may assist in COVID-19 management. However, existing image segmentation using AI suffers from a lack of accuracy and explainability, which prevents its adoption in actual clinical practice. In this paper, we investigated an attention-based image segmentation method for COVID-19 CT imaging with enhanced interpretation capabilities. Specifically, we developed U-Net architecture-based for segmentation with attention coefficients to produce a salient feature map. We use the DICE score and accuracy to perform a comprehensive model evaluation. We compared to other well-known methods such as Light U-Net, COPLE-Net, and Res U-Net and demonstrated that attention U-Net is superior for COVID-19 segmentation tasks in terms of performance and explainability. We also developed the tool as a web-application with a graphic user interface with the goal to translate this AI-driven clinical decision-support system for real-world clinical use. © 2022 IEEE.

Introduction of medical genomics and clinical informatics integration for p-Health care

Achieving predictive, precise, participatory, preventive, and personalized health (abbreviated as p-Health) requires comprehensive evaluations of an individual's conditions captured by various measurement technologies. Since the 1950s, analysis of care providers' and physicians' notes and measurement data by computers to improve healthcare delivery has been termed clinical informatics. Since the 2010s, wide adoptions of Electronic Health Records (EHRs) have greatly improved clinical informatics development with fast growing pervasive wearable technologies that continuously capture the human physiological profile in-clinic (EHRs) and out-of-clinic (PHRs or Personal Health Records) to bolster mobile health (mHealth). In addition, after the Human Genome Project in the 1990s, medical genomics has emerged to capture the high-throughput molecular profile of a person. As a result, integrated data analytics is becoming one of the fast-growing areas under Biomedical Big Data to improve human healthcare outcomes. In this chapter, we first introduce the scope of data integration and review applications, data sources, and tools for clinical informatics and medical genomics. We then describe the data integration analytics at the raw data level, feature level, and decision level with case studies, and the opportunity for research and translation using advanced artificial intelligence (AI), such as deep learning. Lastly, we summarize the opportunities in biomedical big data integration that can reshape healthcare toward p-health.

Using natural language processing on free-Text clinical notes to identify patients with long-Term COVID effects

As of May 15th, 2022, the novel coronavirus SARS-COV-2 has infected 517 million people and resulted in more than 6.2 million deaths around the world. About 40% to 87% of patients suffer from persistent symptoms weeks or months after their original infection. Despite remarkable progress in preventing and treating acute COVID-19 conditions, the clinical diagnosis of long-Term COVID remains difficult. In this work, we use free-Text clinical notes and natural language processing (NLP) techniques to explore long-Term COVID effects. We first obtain free-Text clinical notes from 719 outpatient encounters representing patients treated by physicians at Emory Clinic to detect patterns in patients with long-Term COVID symptoms. We apply state-of-The-Art NLP frameworks to automatically identify patients with long-Term COVID effects, achieving 0.881 recall (sensitivity) score for note-level prediction. We further interpret the prediction outcomes and discuss potential phenotypes. Our work aims to provide a data-driven solution to identify patients who have developed persistent symptoms after acute COVID infection. With this work, clinicians may be able to identify patients who have long-Term COVID symptoms to optimize treatment. © 2022 Owner/Author.

Identification of covid-19 severity and associated genetic biomarkers based on scrna-SEQ data

Bio-marker identification for COVID-19 remains a vital research area to improve current and future pandemic responses. Innovative artificial intelligence and machine learning-based systems may leverage the large quantity and complexity of single cell sequencing data to quickly identify disease with high sensitivity. In this study, we developed a novel approach to classify patient COVID-19 infection severity using single-cell sequencing data derived from patient BronchoAlveolar Lavage Fluid (BALF) samples. We also identified key genetic biomarkers associated with COVID-19 infection severity. Feature importance scores from high performing COVID-19 classifiers were used to identify a set of novel genetic biomarkers that are predictive of COVID-19 infection severity. Treatment development and pandemic reaction may be greatly improved using our novel big-data approach. Our implementation is available on https://github.com/aekanshgoel/COVID-19-scRNAseq. © 2022 Owner/Author.

Disease profile and plasma neutralizing activity of post-vaccination Omicron BA.1 infection in Tianjin, China: a retrospective study (preprint)

BackgroundSARS-CoV-2 Omicron variant BA.1 first emerged on the Chinese mainland in January 2022 in Tianjin and caused a large wave of infections. During mass PCR testing, a total of 430 cases infected with Omicron were recorded between January 8 and February 7, 2022, with no new infections detected for the following 16 days. Most patients had been vaccinated with SARS-CoV-2 inactivated vaccines. The disease profile associated with BA.1 infection, especially after vaccination with inactivated vaccines, is unclear. Whether BA.1 breakthrough infection after receiving inactivated vaccine could create a strong enough humoral immunity barrier against Omicron is not yet investigated. MethodsWe collected the clinical information and vaccination history of the 430 COVID-19 patients infected with Omicron BA.1. Re-positive cases and inflammation markers were monitored during the patients convalescence phase. Ordered multiclass logistic regression model was used to identify risk factors for COVID-19 disease severity. Authentic virus neutralization assays against SARS-CoV-2 wildtype, Beta and Omicron BA.1 were conducted to examine the plasma neutralizing titers induced after post-vaccination Omicron BA.1 infection, and were compared to a group of uninfected healthy individuals who were selected to have a matched vaccination profile. FindingsAmong the 430 patients, 316 (73.5%) were adults with a median age of 47 years, and 114 (26.5%) were under-age with a median age of 10 years. Female and male patients account for 55.6% and 44.4%, respectively. Most of the patients presented with mild (47.7%) to moderate diseases (50.2%), with only 2 severe cases (0.5%) and 7 (1.6%) asymptomatic infections. No death was recorded. 341 (79.3%) of the 430 patients received inactivated vaccines (54.3% BBIBP-CorV vs. 45.5% CoronaVac), 49 (11.4%) received adenovirus-vectored vaccines (Ad5-nCoV), 2 (0.5%) received recombinant protein subunit vaccines (ZF2001), and 38 (8.8%) received no vaccination. No vaccination is associated with a substantially higher ICU admission rate among Omicron BA.1 infected patients (2.0% for vaccinated patients vs. 23.7% for unvaccinated patients, P<0.001). Compared with adults, child patients presented with less severe illness (82.5% mild cases for children vs. 35.1% for adults, P<0.001), no ICU admission, fewer comorbidities (3.5% vs. 53.2%, P<0.001), and less chance of turning re-positive on nucleic acid tests (12.3% vs. 22.5%, P=0.019). For adult patients, compared with no prior vaccination, receiving 3 doses of inactivated vaccine was associated with significantly lower risk of severe disease (OR 0.227 [0.065-0.787], P=0.020), less ICU admission (OR 0.023 [0.002-0.214], P=0.001), lower re-positive rate on PCR (OR 0.240 [0.098-0.587], P=0.002), and shorter duration of hospitalization and recovery (OR 0.233 [0.091-0.596], P=0.002). At the beginning of the convalescence phase, patients who had received 3 doses of inactivated vaccine had substantially lower systemic immune-inflammation index (SII) and C-reactive protein than unvaccinated patients, while CD4+/CD8+ ratio, activated Treg cells and Th1/Th2 ratio were higher compared to their 2-dose counterparts, suggesting that receipt of 3 doses of inactivated vaccine could step up inflammation resolution after infection. Plasma neutralization titers against Omicron, Beta, and wildtype significantly increased after breakthrough infection with Omicron. Moderate symptoms were associated with higher plasma neutralization titers than mild symptoms. However, vaccination profiles prior to infection, whether 2 doses versus 3 doses or types of vaccines, had no significant effect on post-infection neutralization titer. Among recipients of 3 doses of CoronaVac, infection with Omicron BA.1 largely increased neutralization titers against Omicron BA.1 (8.7x), Beta (4.5x), and wildtype (2.2x), compared with uninfected healthy individuals who have a matched vaccination profile. InterpretationReceipt of 3-dose inactivated vaccines can substantially reduce the disease severity of Omicron BA.1 infection, with most vaccinated patients presenting with mild to moderate illness. Child patients present with less severe disease than adult patients after infection. Omicron BA.1 convalescents who had received inactivated vaccines showed significantly increased plasma neutralizing antibody titers against Omicron BA.1, Beta, and wildtype SARS-CoV-2 compared with vaccinated healthy individuals. FundingThis research is supported by Changping Laboratory (CPL-1233) and the Emergency Key Program of Guangzhou Laboratory (EKPG21-30-3), sponsored by the Ministry of Science and Technology of the Peoples Republic of China. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSPrevious studies (many of which have not been peer-reviewed) have reported inconsistent findings regarding the effect of inactivated vaccines against the Omicron variant. On Mar 6, 2022, we searched PubMed with the query "(SARS-CoV-2) AND ((Neutralisation) OR (Neutralisation)) AND ((Omicron) OR (BA.1)) AND (inactivated vaccine)", without date or language restrictions. This search identified 18 articles, of which 13 were directly relevant. Notably, the participants in many of these studies have received only one or two doses of inactivated vaccine with heterologous booster vaccination; other studies have a limited number of participants receiving inactivated vaccines. Added value of this studyTo date, this is the first study to report on the protective effect of inactivated vaccines against the severe disease caused by the Omicron variant. We examine and compare the disease profile of adults and children. Furthermore, we estimate the effect of post-vaccination omicron infection on plasma neutralization titers against Omicron and other SARS-COV-2 variants. Specifically, the disease profile of Omicron convalescents who had received two-dose primary series of inactivated vaccines with or without a booster dose prior to infection is compared with unvaccinated patients. We also analyzed the effect of infection on neutralizing activity by comparing vaccinated convalescents with vaccinated healthy individuals with matched vaccination profiles. Implications of all the available evidenceCompared with adults, child patients infected with Omicron tend to present with less severe disease and are less likely to turn re-positive on nucleic acid tests. Receipt of two-dose primary series or three doses of inactivated vaccine is a protective factor against severe disease, ICU admission, re-positive PCR and longer hospitalization. The protection afforded by a booster dose is stronger than two-dose primary series alone. Besides vaccination, infection with Omicron is also a key factor for elevated neutralizing antibody titers, enabling cross-neutralization against Omicron, wildtype (WT) and the Beta variant.

Learning from heterogeneous data via contrastive learning: An application in multi-source covid-19 radiography

The ongoing COVID-19 pandemic has overloaded current healthcare systems, including radiology systems and departments. Machine learning-based medical imaging diagnostic approaches play an important role in tracking the spread of this virus, identifying high-risk patients, and controlling infections in real-time. Researchers aggregate radiographic samples from different data sources to establish a multi-source learning scheme to mitigate the insufficiency of COVID-19 samples from individual hospitals, especially in the early stage of the disease. However, data heterogeneity across different clinical centers with various imaging conditions is considered a significant limitation in model performance. This paper proposes a contrastive learning scheme for the automatic diagnosis of COVID-19 to effectively mitigate data heterogeneity in multi-source data and learn a robust and generalizable model. Inspired by advances in domain adaptation, we employ contrastive training objectives to promote intra-class cohesion across different data sources and inter-class separation of infected and non-infected cases. Extensive experiments on two public COVID-19 CT datasets demonstrate the effectiveness of the proposed method for tackling data heterogeneity problems with boosted diagnosis performance. Moreover, benefiting from the contrastive learning framework, our method can be generalized to solve data heterogeneity problems under a broader multi-source learning setting. © 2021 IEEE

Automated risk assessment of covid-19 patients at diagnosis using electronic healthcare records

COVID-19 causes significant morbidity and mortality and early intervention is key to minimizing deadly complications. Available treatments, such as monoclonal antibody therapy, may limit complications, but only when given soon after symptom onset. Unfortunately, these treatments are often expensive, in limited supply, require administration within a hospital setting, and should be given before the onset of severe symptoms. These challenges have created the need for early triage of patients likely to develop life-threatening complications. To meet this need, we developed an automated patient risk assessment model using a real-world hospital system dataset with over 17,000 COVID-positive patients. Specifically, for each COVID-positive patient, we generate a separate risk score for each of four clinical outcomes including death within 30 days, mechanical ventilator use, ICU admission, and any catastrophic event (a superset of dangerous outcomes). We hypothesized that a deep learning binary classification approach can generate these four risk scores from electronic healthcare records data at the time of diagnosis. Our approach achieves significant performance on the four tasks with an area under receiver operating curve (AUROC) for any catastrophic outcome, death within 30 days, ventilator use, and ICU admission of 86.7%, 88.2%, 86.2%, and 87.8%, respectively. In addition, we visualize the sensitivity and specificity of these risk scores to allow clinicians to customize their usage within different clinical outcomes. We believe this work fulfills a clear clinical need for early detection of objective clinical outcomes and can be used for early screening for treatment intervention. © 2021 IEEE

COVID-19 diagnosis using model agnostic meta-learning on limited chest X-ray images

In the past year, detection of Coronavirus infection has demonstrated itself to be a challenging task. The gold standard for detection, real-time reverse transcription polymerase chain reaction (RT-PCR) testing, has a few shortcomings, including high false negative rates, long turn-around times, and limited availability. Applying machine learning for automatic analysis on chest X-rays can overcome these issues, but the limited amount of data with which to train inhibits development of robust deep neural networks. In this paper, we demonstrate the feasibility of performing few-shot learning to classify COVID-19 chest X-rays by utilizing a Model-Agnostic Meta-Learning (MAML) algorithm. We compare the improved variant of MAML, named MAML++, to other state-of-the-art machine learning strategies and demonstrate the robust and superior performance in classification accuracy. In addition, we explore the effect of the number of images made available to the sub-learners used for training MAML++ and show that increasing the number of images leads to diminishing returns in performance. Lastly, we compare MAML++ to the original MAML algorithm and discuss the shortcomings of MAML-based algorithms in classification problems. © 2021 ACM.

COVID-19 cases from the first local outbreak of SARS-CoV-2 B.1.1.7 variant in China presented more serious clinical features: a prospective, comparative cohort study (preprint)

Background: The SARS-CoV-2 B.1.1.7 variant which was first identified in the United Kingdom (U.K.) has increased sharply in numbers worldwide and was reported to be more contagious. On January 17, 2021, a COVID-19 clustered outbreak caused by B.1.1.7 variant occurred in a community in Daxing District, Beijing, China. Three weeks prior, another non-variant (lineage B.1.470) COVID-19 outbreak occurred in Shunyi District, Beijing. This study aimed to investigate the clinical features of B.1.1.7 variant infection. Methods: A prospective cohort study was conducted on COVID-19 cases admitted to Ditan hospital since January 2020. Data of 74 COVID-19 cases from two independent COVID-19 outbreaks in Beijing were extracted as study subjects from a Cloud Database established in Ditan hospital, which included 41 Shunyi cases (Shunyi B.1.470 group) and 33 Daxing cases (Daxing B.1.1.7 group) that have been hospitalized since December 25, 2020 and January 17, 2021, respectively. We conducted a comparison of the clinical characteristics, RT-qPCR results and genomic features between the two groups. Findings: Cases from Daxing B.1.1.7 group (15 [45.5%] male; median age, 39 years [range, 30.5, 62.5]) and cases from Shunyi B.1.470 group (25 [61.0%] male; median age, 31 years [range, 27.5, 41.0]) had a statistically significant difference in median age (P =0.014). Seven clinical indicators of Daxing B.1.1.7 group were significantly higher than Shunyi B.1.470 group including patients having fever over 38 (14/33 [46.43%] in Daxing B.1.1.7 group vs. 9/41 (21.95%) in Shunyi B.1.470 group [P = 0 .015]), C-reactive protein ([CRP, mg/L], 4.30 [2.45, 12.1] vs. 1.80, [0.85, 4.95], [P = 0.005]), Serum amyloid A ([SAA, mg/L], 21.50 [12.50, 50.70] vs. 12.00 [5.20, 26.95], [P = 0.003]), Creatine Kinase ([CK, U/L]), 110.50 [53.15,152.40] vs. 70.40 [54.35,103.05], [P = 0.040]), D-dimer ([DD, mg/L], 0.31 [0.20, 0.48] vs. 0.24 [0.17,0.31], [P = 0.038]), CD4+ T lymphocyte ([CD4+ T, mg/L], [P = 0.003]) , and Ground-glass opacity (GGO) in lung (15/33 [45.45%] vs. 5/41 [12.20%], [P =0.001]). After adjusting for the age factor, B.1.1.7 variant infection was the risk factor for CRP (P = 0.045, Odds ratio [OR] 2.791, CI [1.025, 0.8610]), SAA (0.011, 5.031, [1.459, 17.354]), CK (0.034, 4.34, [0.05, 0.91]), CD4+ T ( 0.029, 3.31, [1.13, 9.71]), and GGO (0.005, 5.418, [1.656, 17.729]) of patients. The median Ct value of RT-qPCR tests of the N-gene target in the Daxing B.1.1.7 group was significantly lower than the Shunyi B.1.470 group (P=0.036). The phylogenetic analysis showed that only 2 amino acid mutations in spike protein were detected in B.1.470 strains while B.1.1.7 strains had 3 deletions and 7 mutations. Interpretation: Clinical features including a more serious inflammatory response, pneumonia and a possible higher viral load were detected in the cases infected with B.1.1.7 SARS-CoV-2 variant. It could therefore be inferred that the B.1.1.7 variant may have increased pathogenicity.

EXAM: An Explainable Attention-based Model for COVID-19 Automatic Diagnosis

The ongoing coronavirus disease 2019 (COVID-19) is still rapidly spreading and has caused over 7,000,000 infection cases and 400,000 deaths around the world. To come up with a fast and reliable COVID-19 diagnosis system, people seek help from machine learning area to establish computer-Aided diagnosis systems with the aid of the radiological imaging techniques, like X-ray imaging and computed tomography imaging. Although artificial intelligence based architectures have achieved great improvements in performance, most of the models are still seemed as a black box to researchers. In this paper, we propose an Explainable Attention-based Model (EXAM) for COVID-19 automatic diagnosis with convincing visual interpretation. We transform the diagnosis process with radiological images into an image classification problem differentiating COVID-19, normal and community-Acquired pneumonia (CAP) cases. Combining channel-wise and spatial-wise attention mechanism, the proposed approach can effectively extract key features and suppress irrelevant information. Experiment results and visualization indicate that EXAM outperforms recent state-of-Art models and demonstrate its interpretability. © 2020 ACM.

55-Day-Old Female Infant infected with COVID 19: presenting with pneumonia, liver injury, and heart damage ;The Journal of Infectious Diseases ;Oxford Academic

Previous studies on the pneumonia outbreak caused by the 2019 novel coronavirus disease (COVID-19) were mainly based on information from adult populations. Limited data are available for children with COVID-19, especially for infected infants. We report a 55-day-old case with COVID-19 confirmed in China and describe the identification, diagnosis, clinical course, and treatment of the patient, including the disease progression from day 7 to day 11 of illness. This case highlights that children with COVID-19 can also present with multiple organ damage and rapid disease changes. When managing such patients, frequent and careful clinical monitoring is essential.

Clinical Features of 69 Cases with Coronavirus Disease 2019 in Wuhan, China ;Clinical Infectious Diseases ;Oxford Academic

Abstract Background From December 2019 to February 2020, 2019 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a serious outbreak of coronavirus disease 2019 (COVID-19) in Wuhan, China. Related clinical features are needed. Methods We reviewed 69 patients who were hospitalized in Union hospital in Wuhan between January 16 to January 29, 2020. All patients were confirmed to be infected with SARS-CoV-2 and the final date of follow-up was February 4, 2020. Results The median age of 69 enrolled patients was 42.0 years (IQR 35.0-62.0), and 32 patients (46%) were men. The most common symptoms were fever (60[87%]), cough (38[55%]), and fatigue (29[42%]). Most patients received antiviral therapy (66 [98.5%] of 67 patients) and antibiotic therapy (66 [98.5%] of 67 patients). As of February 4, 2020, 18 (26.9%) of 67 patients had been discharged, and five patients had died, with a mortality rate of 7.5%. According to the lowest SpO2 during admission, cases were divided into the SpO2≥90% group (n=55) and the SpO2<90% group (n=14). All 5 deaths occurred in the SpO2<90% group. Compared with SpO2≥90% group, patients of the SpO2<90% group were older, and showed more comorbidities and higher plasma levels of IL6, IL10, lactate dehydrogenase, and c reactive protein. Arbidol treatment showed tendency to improve the discharging rate and decrease the mortality rate. Conclusions COVID-19 appears to show frequent fever, dry cough, and increase of inflammatory cytokines, and induced a mortality rate of 7.5%. Older patients or those with underlying comorbidities are at higher risk of death.

Dysregulation of immune response in patients with COVID-19 in Wuhan, China ;Clinical Infectious Diseases ;Oxford Academic

Abstract Background In December 2019, coronavirus disease 2019 (COVID-19) emerged in Wuhan and rapidly spread throughout China. Methods Demographic and clinical data of all confirmed cases with COVID-19 on admission at Tongji Hospital from January 10 to February 12, 2020, were collected and analyzed. The data of laboratory examinations, including peripheral lymphocyte subsets, were analyzed and compared between severe and non-severe patients. Results Of the 452 patients with COVID-19 recruited, 286 were diagnosed as severe infection. The median age was 58 years and 235 were male. The most common symptoms were fever, shortness of breath, expectoration, fatigue, dry cough and myalgia. Severe cases tend to have lower lymphocytes counts, higher leukocytes counts and neutrophil-lymphocyte-ratio (NLR), as well as lower percentages of monocytes, eosinophils, and basophils. Most of severe cases demonstrated elevated levels of infection-related biomarkers and inflammatory cytokines. The number of T cells significantly decreased, and more hampered in severe cases. Both helper T cells and suppressor T cells in patients with COVID-19 were below normal levels, and lower level of helper T cells in severe group. The percentage of naïve helper T cells increased and memory helper T cells decreased in severe cases. Patients with COVID-19 also have lower level of regulatory T cells, and more obviously damaged in severe cases. Conclusions The novel coronavirus might mainly act on lymphocytes, especially T lymphocytes. Surveillance of NLR and lymphocyte subsets is helpful in the early screening of critical illness, diagnosis and treatment of COVID-19.