Accurate classification of COVID‐19 patients with different severity via machine learning

Abstract

Infection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could cause dramatic response in coronavirus disease 2019 (COVID-19) patients at multi-omics level, 1-3 thus it is essential to systematically assess the pathogenesis of COVID-19. In our previous study, we presented the first trans-omics landscape of 236 COVID-19 patients with 4 clinical severity groups (including asymptomatic, mild, severe and critically ill cases) and found that the mild and severe COVID-19 patients shared several similar characteristics. 4 However, it is crucial to discriminate mild from severe COVID-19 patients to prevent the latter from the progression of disease by facilitating early intervention. Herein, we developed an extreme gradient boosting (XGBoost) machine-learning model to predict the COVID-19 severities by leveraging multi-omics data. Briefly, we randomly stratified samples for the training set (80%) and the independent testing set (20%) (Figure 1A, see Methods in the Supporting Information). After normalization, a total of 297 multi-omics features were preliminarily selected by applying a hybrid method (see Methods in the Supporting Information). The XGBoost model was trained on the training set with the preliminarily selected features, achieving mean micro-average AUROC (area under the receiver operating characteristic curve) and mean micro-average AUPR (area under the precisionrecall curve) of 0.9715 (95% CI, 0.9497-0.9932) and 0.9495 (95% CI, 0.9086-0.9904), respectively (Figure 1B and C). This showed strong generalizable discrimination among the four severities based on fivefold cross-validation over 100 iterations. The multi-omics features were prioritized and ranked by the XGBoost model and the SHAP (SHapley Additive exPlanations, see Methods in the Supporting Information) value. Top 60 important features were further selected consisting of 19 proteins, 11 metabolites, 7 lipids, and 23 mRNAs (Figure 1D, Figures S1-S4). With the top 60 features, the XGBoost model was retrained and val-This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Measurement of SARS-CoV-2 Antibody Titers Improves the Prediction Accuracy of COVID-19 Maximum Severity by Machine Learning in Non-Vaccinated Patients

“…The optimum hyperparameters for each model were found by grid search, and are described in Supplemental Table S3. The XGBoost classifies samples into several categories based on a trained gradient boosting decision tree and has been used for similar studies (43)(44)(45). To investigate the impact of antibody titers as features on the model accuracy, models with and without inclusion of the results of antibody testing as input data were built, and the feature importance was calculated.…”

Early and long term antibody kinetics of asymptomatic and mild disease COVID-19 patients

“…Specifically, AI models were designed to predict the prevalence of asymptomatic COVID-19 carriers 18 . However, only limited results are available regarding classification of asymptomatic carriers, and predicting the course of the disease based on antibody kinetics 19 . The aim of the current study is therefore to evaluate early and late antibody kinetics in asymptomatic and mildly symptomatic cases, and to provide further insights into the association between antibody levels and disease phase in a longitudinal household study design.…”

How to translate the knowledge of COVID‐19 into the prevention of Omicron variants

“…For example, inflammatory signatures and machine learning model for accurate classification of COVID‐19 patients with different severity were developed to predict the occurrence of critical illness in patients with COVID‐19. 7 , 8 On basis of the systemic inflammatory panel from multiple clinical centres, the exacerbation of disease in COVID‐19 patients could reliably be predicted 20 days before the occurrence. If these approaches can be applied to patients infected with Omicron variants, clinical management can be directed and prioritized to prevent the progression of the disease and optimize resource utilization.…”