Related Experiment Video
Updated: Jul 9, 2025
05:28
Evaluation of a Reliable Biomarker in a Cecal Ligation and Puncture-Induced Mouse Model of Sepsis
Published on: December 9, 2022
3.5K
A simplified machine learning model utilizing platelet-related genes for predicting poor prognosis in sepsis
Yingying Diao1, Yan Zhao1, Xinyao Li1
1National Clinical Research Center for Laboratory Medicine, Department of Laboratory Medicine, The First Hospital of China Medical University, Shenyang, China.
Frontiers in Immunology
|December 6, 2023
Summary
A machine learning model using platelet genes predicts sepsis prognosis. This model aids early treatment decisions for sepsis patients, improving personalized medicine and patient care.
Area of Science:
- Genomics
- Machine Learning
- Sepsis Research
Background:
- Thrombocytopenia is a known prognostic factor in sepsis.
- The link between platelet-related genes and sepsis outcomes requires further investigation.
- Machine learning offers a novel approach to predict sepsis prognosis.
Purpose of the Study:
- To develop and validate a machine learning model for predicting poor prognosis in sepsis using platelet-related genes.
- To assess the predictive performance of the model against established clinical scores.
- To explore the potential of gene-based models in personalized sepsis management.
Main Methods:
- Retrospective analysis of platelet data from 365 sepsis patients.
- Identification of key platelet-related genes using COX analysis, LASSO, and SVM.
- Training and validation of a machine learning model on six diverse platforms (719 patients).
- Comparison of model performance with Acute Physiology and Chronic Health Evaluation II (APACHE II) and Sequential Organ-Failure Assessment (SOFA) scores.
Main Results:
- A platelet count below 100×10^9/L independently increased the risk of death in sepsis patients (OR = 2.523).
- The machine learning model, based on five platelet-related genes, achieved area under the curve (AUC) values from 0.5 to 0.795 across platforms.
- On one platform, the model (AUC=0.795) outperformed the APACHE II score (AUC=0.761), with improved performance (AUC=0.812) when age was included.
- On another platform, the model's AUC improved from 0.5 to 0.583 with age inclusion, compared to APACHE II (0.604) and SOFA (0.542).
Conclusions:
- The developed machine learning model demonstrates broad applicability for early treatment decisions in sepsis.
- This gene-based predictive model holds promise for advancing personalized medicine in sepsis care.
- The findings support the integration of genomic data into clinical decision-making for improved patient outcomes.
