doi: 10.1002/ctm2.70241.
Machine-learning analysis identifies "elite" viral controllers with increased survival and homeostatic responses in critical COVID-19
Affiliations
- PMID: 40285348
- PMCID: PMC12031886
- DOI: 10.1002/ctm2.70241
Item in Clipboard
Machine-learning analysis identifies "elite" viral controllers with increased survival and homeostatic responses in critical COVID-19
Clin Transl Med.
2025 May.
No abstract available
Conflict of interest statement
JFBM, AT, FB, RA, JME and APT have a patent application on SARS‐CoV‐2 antigenemia as a predictor of mortality in COVID‐19.
The remaining authors declare no conflicts of interest.
Figures
SHAP (SHapley Additive exPlanations) value distribution of the top 20 features obtained in the machine learning model for identifying the 90‐day mortality. Each point is a patient. The horizontal position of each point represents the SHAP value (importance higher or lower for prediction) and the direction of each feature for predicting a particular patient. Positive SHAP values predict true positives, and negative SHAP values predict true negatives. The red colour indicates high values, while the blue colour indicates low values of the features for a specific patient. Plasma biomarker values were ln‐transformed. SHAP values were calculated using the Python package XGBoost. Abbreviations: SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2; IL, interleukin; RANTES, regulated on activation, normal T cell expressed and secreted protein; IgG, immunoglobulin G; IgM, immunoglobulin M; TREM‐1, triggering receptor expressed on myeloid cells 1; CCL2, chemokine (C‐C motif) ligand 2; CD27, cluster differentiation 27 molecule; SP‐D, Surfactant protein D; PTX‐3, pentraxin 3; CXCL10, C‐X‐C motif chemokine ligand 10; VCAM‐1, vascular cell adhesion protein.
90‐day mortality groups (A) and partitional clustering groups (B) visualized using t‐Distributed Stochastic Neighbor Embedding. (C) Kaplan–Meier curves show the cumulative probability of mortality in the first 90 days following ICU admission. T‐SNE, t‐Distributed Stochastic Neighbor Embedding, p‐value, level of significance.
Heat map showing the three distinct patterns of inflammatory immune response by 90‐day mortality risk group (combitypes). Abbreviations: SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2; IgM, immunoglobulin M; IgG, immunoglobulin G; RANTES, regulated on activation normal T‐cell expressed and secreted protein; TREM‐1, triggering receptor expressed on myeloid cells; G‐CSF, granulocyte colony‐stimulating factor; ICAM‐1, intercellular adhesion molecule 1; VCAM‐1, vascular cell adhesion molecule 1; IL, interleukin; TNF, tumour necrosis factor; CXCL‐10, C‐X‐C motif chemokine ligand 10; CCL2, chemokine (C‐C motif) ligand 2; IFN, interferon; PD‐L1, programmed death‐ligand 1; CD‐27, cluster differentiation 27 molecule; SP‐D, surfactant protein D.
Machine learning analysis combining host‐response and virological data improves the characterization of subgroups of critically ill COVID‐19 patients with different prognosis. Abbreviations: SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2; IgM, immunoglobulin M; IgG, immunoglobulin G; RANTES, regulated on activation normal T‐cell expressed and secreted protein; TREM‐1, triggering receptor expressed on myeloid cells; G‐CSF, granulocyte colony‐stimulating factor; ICAM‐1, intercellular adhesion molecule 1; VCAM‐1, vascular cell adhesion molecule 1; IL, interleukin; TNF, tumour necrosis factor; CXCL‐10, C‐X‐C motif chemokine ligand 10; CCL2, chemokine (C‐C motif) ligand 2; IFN, interferon; PD‐L1, programmed death‐ligand 1; CD‐27, cluster differentiation 27 molecule; SP‐D, surfactant protein D.
References
Grants and funding
LinkOut - more resources
Full Text Sources