Specific diagnostic model for bacterial pneumonia constructed by combining multiple omics and multi machine learning...

Area of Science:

• Biochemistry
• Genomics
• Machine Learning

Background:

• Accurate diagnosis of bacterial pneumonia (BPs) is challenging, necessitating novel biomarkers.
• Distinguishing BPs from non-bacterial pneumonia (NBPs) is crucial for effective treatment.

Purpose of the Study:

• To discover specific biomarkers for bacterial pneumonia (BPs).
• To develop a machine learning-based diagnostic model for BPs.

Main Methods:

• Serum samples from 45 BPs and 35 NBPs patients underwent miRNA, proteomic, and metabolomic sequencing.
• Eight potential biomarkers were identified through profiling.
• Six machine learning algorithms (LR, RF, SVM, XGBoost, LightGBM, ExtraTree) were employed for model development.

Main Results:

• Eight candidate biomarkers were screened for their diagnostic potential in BPs.
• The logistic regression (LR) model achieved the highest AUC of 0.892.
• The LR model exhibited a sensitivity of 0.769 and specificity of 0.900 for BPs diagnosis.

Conclusions:

• This research identified 8 potential biomarkers for bacterial pneumonia (BPs) diagnosis.
• A reliable diagnostic model using logistic regression was established for BPs.