Interpretable Ensemble Learning Predicts Antibiotic Resistance in Treponema denticola Using Expert Classifiers
Introduction and Objectives: Antibiotic resistance is a global health concern, contributing to prolonged hospital stays, increased medical costs, and higher mortality rates. Addressing antimicrobial resistance (AMR) in periodontal infections requires targeted therapies and a multifaceted approach. T...
Sábháilte in:
| Príomhchruthaitheoirí: | , , |
|---|---|
| Formáid: | Alt |
| Teanga: | Béarla |
| Foilsithe / Cruthaithe: |
Elsevier
2025-10-01
|
| Sraith: | International Dental Journal |
| Ábhair: | |
| Rochtain ar líne: | http://www.sciencedirect.com/science/article/pii/S002065392500173X |
| Clibeanna: |
Cuir clib leis
Níl clibeanna ann, Bí ar an gcéad duine le clib a chur leis an taifead seo!
|
| Achoimre: | Introduction and Objectives: Antibiotic resistance is a global health concern, contributing to prolonged hospital stays, increased medical costs, and higher mortality rates. Addressing antimicrobial resistance (AMR) in periodontal infections requires targeted therapies and a multifaceted approach. This study aims to predict and classify AMR genomic sequences in Treponema denticola, a key pathogen in periodontal disease, using machine learning (ML). Methods: UniProt FASTA sequences were used to investigate AMR in T. denticola. Data were retrieved and preprocessed using the BioPython library in a Jupyter Notebook. A structured approach included data exploration, feature extraction, and visualization. Four classification models – Random Forest, Support Vector Machine (SVM), Gradient Boosting, and Neural Network (Multilayer Perceptron Classifier) – were optimized using specific hyperparameters. Model performance was evaluated using fivefold stratified cross-validation. A Voting Classifier, combining multiple models, was implemented to enhance predictive accuracy. Results: The Voting Classifier outperformed Random Forest, SVM, Gradient Boosting, and Neural Network models, achieving the highest test accuracy (96.46%) and F1-score (0.9646). High accuracy was also demonstrated by SVM and Neural Networks (95.58%), but the robustness of the Voting Classifier was highlighted by its ability to balance accuracy with low log loss (0.1504). Conclusion: This study highlights the effectiveness of the Voting Classifier in classifying AMR genomic sequences in T. denticola. The findings underscore the potential of interpretable ML approaches for advancing AMR research in periodontal pathogens and informing targeted therapeutic strategies. Clinical Relevance: The ability to accurately predict AMR in T. denticola using ML models like the Voting Classifier can significantly enhance clinical decision-making. By identifying resistance patterns, clinicians can tailor antibiotic therapies more effectively, reducing treatment failures and mitigating the spread of resistance. This approach also supports the development of novel antimicrobial agents and strengthens public health surveillance efforts, particularly in resource-limited settings where periodontal infections are prevalent. |
|---|---|
| ISSN: | 0020-6539 |