Scholarlink

Objective <p>To explore the application value of a multi-omics fusion model based on clinical indicators and genomic features in the prediction of antibiotic resistance in Gram-negative bacterial pneumonia, screen key biomarkers influencing resistance, construct and validate an efficient prediction model, and provide a basis for the precise clinical use of antibiotics.</p> Methods <p>A total of 836 patients diagnosed with Gram-negative bacterial pneumonia and receiving antibiotic treatment in the hospital from January 2023 to June 2025 were selected as the study subjects. They were randomly divided into a training set (585 cases) and a validation set (251 cases) at a ratio of 7:3. General clinical data, laboratory test indicators, and genomic feature data of the patients were collected. Key predictive indicators were screened through univariate analysis and multivariate logistic regression analysis. Random forest (RF), logistic regression (LR), and extreme gradient boosting (XGBoost) prediction models were constructed using Python 3.9.0 software and related libraries. The performance of the models was evaluated using accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (AUC). The clinical applicability of the models was verified through calibration curves and decision curves to provide a basis for early precise drug use.</p> Results <p>Multivariate analysis showed that length of hospital stay, history of β-lactam antibiotic use, β-lactamase gene (blaKPC), deletion of porin protein gene (ompK36), procalcitonin (PCT), sputum culture colony count, and the situation of combined use of antibacterial drugs were independent predictive factors affecting antibiotic resistance in Gram-negative bacterial pneumonia (<i>P</i> &lt; 0.05). Among the constructed models, the XGBoost model had the best performance, with AUC values of 0.853 and 0.814 in the training set and validation set respectively. Its precision, accuracy, recall, F1 score and other indicators were also superior to those of other models.</p> Conclusion <p>A machine-learning model integrating clinical and genomic features can effectively predict antibiotic resistance in Gram-negative bacterial pneumonia. The XGBoost model performs best, which can provide a reference for formulating individualized anti-infection regimens in the early clinical stage, reduce the spread of drug-resistant bacteria, and improve patient prognosis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning prediction of antibiotic resistance in Gram-negative pneumonia: a clinical-genomic integrated model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月9日

查看全文
  • Jing Guan,
  • Zhaoming Chen,
  • Haoru Wang,
  • Weilong Li,
  • Feng Ye,
  • Shaoqiang Li

摘要

Objective

To explore the application value of a multi-omics fusion model based on clinical indicators and genomic features in the prediction of antibiotic resistance in Gram-negative bacterial pneumonia, screen key biomarkers influencing resistance, construct and validate an efficient prediction model, and provide a basis for the precise clinical use of antibiotics.

Methods

A total of 836 patients diagnosed with Gram-negative bacterial pneumonia and receiving antibiotic treatment in the hospital from January 2023 to June 2025 were selected as the study subjects. They were randomly divided into a training set (585 cases) and a validation set (251 cases) at a ratio of 7:3. General clinical data, laboratory test indicators, and genomic feature data of the patients were collected. Key predictive indicators were screened through univariate analysis and multivariate logistic regression analysis. Random forest (RF), logistic regression (LR), and extreme gradient boosting (XGBoost) prediction models were constructed using Python 3.9.0 software and related libraries. The performance of the models was evaluated using accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (AUC). The clinical applicability of the models was verified through calibration curves and decision curves to provide a basis for early precise drug use.

Results

Multivariate analysis showed that length of hospital stay, history of β-lactam antibiotic use, β-lactamase gene (blaKPC), deletion of porin protein gene (ompK36), procalcitonin (PCT), sputum culture colony count, and the situation of combined use of antibacterial drugs were independent predictive factors affecting antibiotic resistance in Gram-negative bacterial pneumonia (P < 0.05). Among the constructed models, the XGBoost model had the best performance, with AUC values of 0.853 and 0.814 in the training set and validation set respectively. Its precision, accuracy, recall, F1 score and other indicators were also superior to those of other models.

Conclusion

A machine-learning model integrating clinical and genomic features can effectively predict antibiotic resistance in Gram-negative bacterial pneumonia. The XGBoost model performs best, which can provide a reference for formulating individualized anti-infection regimens in the early clinical stage, reduce the spread of drug-resistant bacteria, and improve patient prognosis.