key: cord-0943428-e4aha2i8 authors: wen, shaoqing; Wang, Yi title: Monitoring and predicting viral dynamics in SARS-CoV-2-infected Patients date: 2020-04-17 journal: nan DOI: 10.1101/2020.04.14.20060491 sha: 70b4f74030be481eb8d8e3da7b8620edf78d4e98 doc_id: 943428 cord_uid: e4aha2i8 This manuscript is based on the a simple but robust model we developed urgently to accurately monitor and predict viral dynamics for each SARS-CoV-2-infected patient, given the limited number of RT-PCR tests and the complexity of each individual's physical health situation. In this study, we used the mathematical model to monitor and predict the changes of viral loads from different nasal and throat swab of clinical specimens collected from diagnosed patients. We also tested our real time model by using the data from the SARS-CoV-2-infected patients with different severity. By using this personal model, we can predict the viral dynamics of patients, minimize false-negative test results, and screen the patients who are at risk of testing positive again after recovery. We sincerely thank those who are on the front lines battling SARS-CoV-2 virus. We hope this model will be useful for SARS-CoV-2-infected patients. SARS-CoV-2 viral dynamics revealed that:1) the viral load in lower respiratory tract (LRT; such as sputum) is generally higher than upper respiratory tract (URT; such as nasopharyngeal and oropharyngeal) [1] and, in the upper respiratory tract, higher viral loads are detected in the nose rather than in the throat [2] ; 2) the viral loads in SARS-CoV-2-infected patients peaked soon after the onset of symptoms, resembling influenza-infected patients and distinct from SARS-CoV-infected patients [1] [2] [3] [4] ; 3) viral loads detected in asymptomatic or minimally symptomatic patients indicate the transmission potential in the early phase of infection [2] . However, there are many details that remain puzzled. For example, 1) some patients have presented positive chest CT findings showing multifocal ground-glass changes, but they had negative RT-PCR results at that time [5] ; 2) some patients have rapid changes in viral loads during the window of infection, with the RT-PCR results sometimes positive and sometimes negative [2] [3] [4] ; 3) few recovered patients (absence of clinical symptoms and radiological abnormalities and . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.14.20060491 doi: medRxiv preprint results again 1 or 2 weeks after discharge from hospital [6] . Importantly, it is impossible to continuously collect samples from each SARS-CoV-2-infected patient over the course of the treatment and rehabilitation. In this context, a simple but robust model is expected to accurately monitor and predict viral dynamics for each SARS-CoV-2-infected patient, given the limited number of RT-PCR tests and the complexity of each individual's physical health situation. First of all, we reformat the original sparse matrix drawn from ref [2] (viral loads of 18 Chinese infected patients with many missing data points) into a dense matrix (training set) as follow: We apply a linear regression to the above data format. To build a personal model with few data points, we borrow information from the population. We replicate the data points of the query person N times, where N is the number of population data points. We then concatenate the N folds of personal data points with the population data points to form a combined dataset. In this way, population data points are relatively down weights and act as a prior. The combined dataset is fed into a linear regression model to predict the response value: Cycle threshold (Ct). Ct values of SARS-CoV-2-specific gene on RT-PCR assay are inversely related to viral RNA copy number. Note that nasal data and throat data are also combined with an indicator URT in order to maximize the number of data points. Subsequently, we tested the personal model with another two sets of published data [3, 4] . author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.14.20060491 doi: medRxiv preprint curves, that could be because the treatment with lopinavir-ritonavir, or, more likely, because the occurrence of false-negative test results. Moreover, we should pay special attention to those who have negative RT-PCR results but lower predicted Ct values on day 21 after symptom onset, such as patient 2 and 7, which implies they may turn positive again. In this study, we used a mathematical model to monitor and predict the changes of viral loads from different nasal and throat swab of clinical specimens collected from diagnosed patients. We also tested our real time model by using the data from the SARS-CoV-2-infected patients with different severity. By using this personal model, we can predict the viral dynamics of patients, minimize false-negative test results, and screen the patients who are at risk of testing positive again after recovery. The following factors can optimize our model: 1) more data regarding viral loads of the infected patients; 2) detailed information about clinical features and laboratory results; 3) more samples covering the clinical course, especially key time points (exposure, symptom onset, 5-6 days after symptom onset, recovery, and two weeks after recovery); 4) more attention to the early phase between exposure and symptom onset, which is currently a knowledge gap for SARS-CoV-2 viral dynamics studies. Finally, we hope this model will be useful for SARS-CoV-2-infected patients. Table S1 . Clinical course and viral loads (tested Ct values and predicted Ct values) of 20 SARS-CoV-2-infected patients according to the timeline from symptom onset. . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.14.20060491 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.14.20060491 doi: medRxiv preprint Viral load of SARS-CoV-2 in clinical samples SARS-CoV-2 Viral Load in Upper Respiratory Specimens of Infected Patients Viral Load Kinetics of SARS-CoV-2 Infection in First Two Patients in Korea Epidemiologic Features and Clinical Course of Patients Infected With SARS-CoV-2 in Singapore Chest CT for Typical 2019-nCoV Pneumonia: Relationship to Negative RT-PCR Testing Positive RT-PCR Test Results in Patients Recovered From COVID-19 We sincerely thank those who are on the front lines battling SARS-CoV-2 virus. The authors declare that they have no conflict of interest.