A Comparison of XGBoost, Random Forest, and Nomograph for the Prediction of Disease Severity in Patients With COVID-19 Pneumonia: Implications of Cytokine and Immune Cell Profile.

Background and Aims: The aim of this study was to apply machine learning models and a nomogram to differentiate critically ill from non-critically ill COVID-19 pneumonia patients. Methods: Clinical symptoms and signs, laboratory parameters, cytokine profile, and immune cellular data of 63 COVID-19 pneumonia patients were retrospectively reviewed. Outcomes were followed up until Mar 12, 2020. A logistic regression function (LR model), Random Forest, and XGBoost models were developed. The performance of these models was measured by area under receiver operating characteristic curve (AUC) analysis. Results: Univariate analysis revealed that there was a difference between critically and non-critically ill patients with respect to levels of interleukin-6, interleukin-10, T cells, CD4+ T, and CD8+ T cells. Interleukin-10 with an AUC of 0.86 was most useful predictor of critically ill patients with COVID-19 pneumonia. Ten variables (respiratory rate, neutrophil counts, aspartate transaminase, albumin, serum procalcitonin, D-dimer and B-type natriuretic peptide, CD4+ T cells, interleukin-6 and interleukin-10) were used as candidate predictors for LR model, Random Forest (RF) and XGBoost model application. The coefficients from LR model were utilized to build a nomogram. RF and XGBoost methods suggested that Interleukin-10 and interleukin-6 were the most important variables for severity of illness prediction. The mean AUC for LR, RF, and XGBoost model were 0.91, 0.89, and 0.93 respectively (in two-fold cross-validation). Individualized prediction by XGBoost model was explained by local interpretable model-agnostic explanations (LIME) plot. Conclusions: XGBoost exhibited the highest discriminatory performance for prediction of critically ill patients with COVID-19 pneumonia. It is inferred that the nomogram and visualized interpretation with LIME plot could be useful in the clinical setting. Additionally, interleukin-10 could serve as a useful predictor of critically ill patients with COVID-19 pneumonia.

A Comparison of XGBoost, Random Forest, and Nomograph for the Prediction of Disease Severity in Patients With COVID-19 Pneumonia: Implications of Cytokine and Immune Cell Profile

Background and Aims The aim of this study was to apply machine learning models and a nomogram to differentiate critically ill from non-critically ill COVID-19 pneumonia patients. Methods Clinical symptoms and signs, laboratory parameters, cytokine profile, and immune cellular data of 63 COVID-19 pneumonia patients were retrospectively reviewed. Outcomes were followed up until Mar 12, 2020. A logistic regression function (LR model), Random Forest, and XGBoost models were developed. The performance of these models was measured by area under receiver operating characteristic curve (AUC) analysis. Results Univariate analysis revealed that there was a difference between critically and non-critically ill patients with respect to levels of interleukin-6, interleukin-10, T cells, CD4+ T, and CD8+ T cells. Interleukin-10 with an AUC of 0.86 was most useful predictor of critically ill patients with COVID-19 pneumonia. Ten variables (respiratory rate, neutrophil counts, aspartate transaminase, albumin, serum procalcitonin, D-dimer and B-type natriuretic peptide, CD4+ T cells, interleukin-6 and interleukin-10) were used as candidate predictors for LR model, Random Forest (RF) and XGBoost model application. The coefficients from LR model were utilized to build a nomogram. RF and XGBoost methods suggested that Interleukin-10 and interleukin-6 were the most important variables for severity of illness prediction. The mean AUC for LR, RF, and XGBoost model were 0.91, 0.89, and 0.93 respectively (in two-fold cross-validation). Individualized prediction by XGBoost model was explained by local interpretable model-agnostic explanations (LIME) plot. Conclusions XGBoost exhibited the highest discriminatory performance for prediction of critically ill patients with COVID-19 pneumonia. It is inferred that the nomogram and visualized interpretation with LIME plot could be useful in the clinical setting. Additionally, interleukin-10 could serve as a useful predictor of critically ill patients with COVID-19 pneumonia.

Reverse vaccinology approach for multi-epitope centered vaccine design against delta variant of the SARS-CoV-2.

The ongoing COVID-19 outbreak, initially identified in Wuhan, China, has impacted people all over the globe and new variants of concern continue to threaten hundreds of thousands of people. The delta variant (first reported in India) is currently classified as one of the most contagious variants of SARS-CoV-2. It is estimated that the transmission rate of delta variant is 225% times faster than the alpha variant, and it is causing havoc worldwide (especially in the USA, UK, and South Asia). The mutations found in the spike protein of delta variant make it more infective than other variants in addition to ruining the global efficacy of available vaccines. In the current study, an in silico reverse vaccinology approach was applied for multi-epitope vaccine construction against the spike protein of delta variant, which could induce an immune response against COVID-19 infection. Non-toxic, highly conserved, non-allergenic and highly antigenic B-cell, HTL, and CTL epitopes were identified to minimize adverse effects and maximize the efficacy of chimeric vaccines that could be developed from these epitopes. Finally, V1 vaccine construct model was shortlisted and 3D modeling was performed by refinement, docking against HLAs and TLR4 protein, simulation and in silico expression. In silico evaluation showed that the designed chimeric vaccine could elicit an immune response (i.e., cell-mediated and humoral) identified through immune simulation. This study could add to the efforts of overcoming global burden of COVID-19 particularly the variants of concern.

Critically Ill vs. Non-Critically Ill Patients With COVID-19 Pneumonia: Clinical Features, Laboratory Findings, and Prediction.

Objectives: The objective of this study was to investigate the clinical features and laboratory findings of patients with and without critical COVID-19 pneumonia and identify predictors for the critical form of the disease. Methods: Demographic, clinical, and laboratory data of 63 COVID-19 pneumonia patients were retrospectively reviewed. Laboratory parameters were also collected within 3-5 days, 7-9 days, and 11-14 days of hospitalization. Outcomes were followed up until March 12, 2020. Results: Twenty-two patients developed critically ill pneumonia; one of them died. Upon admission, older patients with critical illness were more likely to report cough and dyspnoea with higher respiration rates and had a greater possibility of abnormal laboratory parameters than patients without critical illness. When compared with the non-critically ill patients, patients with serious illness had a lower discharge rate and longer hospital stays, with a trend towards higher mortality. The interleukin-6 level in patients upon hospital admission was important in predicting disease severity and was associated with the length of hospitalization. Conclusions: Many differences in clinical features and laboratory findings were observed between patients exhibiting non-critically ill and critically ill COVID-19 pneumonia. Non-critically ill COVID-19 pneumonia also needs aggressive treatments. Interleukin-6 was a superior predictor of disease severity.

Comprehensive comparative genomic and microsatellite analysis of SARS, MERS, BAT-SARS, and COVID-19 coronaviruses.

The coronavirus disease 2019 (COVID-19) pandemic has spread around the globe very rapidly. Previously, the evolution pattern and similarity among the COVID-19 causative organism severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and causative organisms of other similar infections have been determined using a single type of genetic marker in different studies. Herein, the SARS-CoV-2 and related ß coronaviruses Middle East respiratory syndrome coronavirus (MERS-CoV), SARS-CoV,  bat coronavirus (BAT-CoV) were comprehensively analyzed using a custom-built pipeline that employed phylogenetic approaches based on multiple types of genetic markers including the whole genome sequences, mutations in nucleotide sequences, mutations in protein sequences, and microsatellites. The whole-genome sequence-based phylogeny revealed that the strains of SARS-CoV-2 are more similar to the BAT-CoV strains. The mutational analysis showed that on average MERS-CoV and BAT-CoV genomes differed at 134.21 and 136.72 sites, respectively, whereas the SARS-CoV genome differed at 26.64 sites from the reference genome of SARS-CoV-2. Furthermore, the microsatellite analysis highlighted a relatively higher number of average microsatellites for MERS-CoV and SARS-CoV-2 (106.8 and 107, respectively), and a lower number for SARS-CoV and BAT-CoV (95.8 and 98.5, respectively). Collectively, the analysis of multiple genetic markers of selected ß viral genomes revealed that the newly born SARS-COV-2 is closely related to BAT-CoV, whereas, MERS-CoV is more distinct from the SARS-CoV-2 than BAT-CoV and SARS-CoV.

Screening of potent phytochemical inhibitors against SARS-CoV-2 protease and its two Asian mutants.

BACKGROUND: COVID-19, declared a pandemic in March 2020 by the World Health Organization is caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). The virus has already killed more than 2.3 million people worldwide. OBJECT: The principal intent of this work was to investigate lead compounds by screening natural product library (NPASS) for possible treatment of COVID-19. METHODS: Pharmacophore features were used to screen a large database to get a small dataset for structure-based virtual screening of natural product compounds. In the structure-based screening, molecular docking was performed to find a potent inhibitor molecule against the main protease (Mpro) of SARS-CoV-2 (PDB ID: 6Y7M). The predicted lead compound was further subjected to Molecular Dynamics (MD) simulation to check the stability of the leads compound with the evolution of time. RESULTS: In pharmacophore-based virtual screening, 2,361 compounds were retained out of 30,927. In the structure-based screening, the lead compounds were filtered based on their docking scores. Among the 2,360 compounds, 12 lead compounds were selected based on their docking score. Kazinol T with NPASS ID: NPC474104 showed the highest docking score of -14.355 and passed criteria of Lipinski's drug-like parameters. Monitoring ADMET properties, Kazinol T showed its safety for consumption. Docking of Kazinol T with two Asian mutants (R60C and I152V) showed variations in binding and energy parameters. Normal mode analysis for ligand-bound and unbound form of protease along with its mutants, revealed displacement and correlation parameters for C-alpha atoms. MD simulation results showed that all ligand-protein complexes remained intact and stable in a dynamic environment with negative Gibbs free energy. CONCLUSIONS: The natural product Kazinol T was a predicted lead compound against the main protease of SARS-CoV-2 and will be the possible treatment for COVID-19.

Vaccine Design from the Ensemble of Surface Glycoprotein Epitopes of SARS-CoV-2: An Immunoinformatics Approach.

The present study aimed to work out a peptide-based multi-epitope vaccine against the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). We predicted different B-cell and T-cell epitopes by using the Immune Epitopes Database (IEDB). Homology modeling of the construct was done using SWISS-MODEL and then docked with different toll-like-receptors (TLR4, TLR7, and TLR8) using PatchDock, HADDOCK, and FireDock, respectively. From the overlapped epitopes, we designed five vaccine constructs C1-C5. Based on antigenicity, allergenicity, solubility, different physiochemical properties, and molecular docking scores, we selected the vaccine construct 1 (C1) for further processing. Docking of C1 with TLR4, TLR7, and TLR8 showed striking interactions with global binding energy of -43.48, -65.88, and -60.24 Kcal/mol, respectively. The docked complex was further simulated, which revealed that both molecules remain stable with minimum RMSF. Activation of TLRs induces downstream pathways to produce pro-inflammatory cytokines against viruses and immune system simulation shows enhanced antibody production after the booster dose. In conclusion, C1 was the best vaccine candidate among all designed constructs to elicit an immune response SARS-CoV-2 and combat the coronavirus disease (COVID-19).