Machine learning-based antibiotic resistance prediction models: An updated systematic review and meta-analysis

Area of Science:

• * Computational biology and bioinformatics
• * Medical informatics and artificial intelligence
• * Infectious disease epidemiology

Background:

• * Escalating antibiotic resistance poses a significant global health threat.
• * Diminishing antibiotic effectiveness necessitates novel prediction strategies.
• * Machine learning (ML) offers promising avenues for predicting bacterial resistance patterns.

Purpose of the Study:

• * To systematically review and quantitatively analyze machine learning models for antibiotic resistance prediction.
• * To evaluate the methodological quality and performance of existing ML models.
• * To provide an up-to-date assessment of research progress in this field.

Main Methods:

• * Comprehensive literature search across PubMed, Embase, and Cochrane Library up to December 2023.
• * Inclusion of studies meeting predefined criteria for ML-based antibiotic resistance prediction.
• * Meta-analysis using a random-effects model and risk of bias assessment.

Main Results:

• * 22 studies (43,628 samples) were reviewed; 10 included in meta-analysis.
• * Random forest, decision trees, and neural networks were common ML algorithms.
• * ML models demonstrated good discriminative ability (AUC=0.78), with high specificity (0.95) but variable sensitivity (0.57).
• * High risk of bias and publication bias were identified in included studies.

Conclusions:

• * Machine learning models show promise for predicting antibiotic resistance.
• * Current evidence suggests good diagnostic performance, but is limited by study quality.
• * Future research requires rigorous design and reporting to improve reliability for clinical use.