Exploring feature limitations in antimicrobial resistance prediction: machine learning and deep learning in A.
Zahra Seraj1, Zahra Ghorbanali2, Fatemeh Zare-Mirakabad3
1Laboratory of Bioinformatics and Drug Design (LBD), Institute of Biochemistry and Biophysics, University of Tehran, Tehran, Iran.
Scientific Reports
|June 28, 2026
Summary
This study develops a computational framework for predicting antimicrobial resistance in Acinetobacter baumannii using gene presence/absence profiles. Data-driven feature selection and deep learning models significantly improve prediction accuracy and specificity for this critical pathogen.
Area of Science:
- Computational biology
- Genomics
- Machine learning
Background:
- Antimicrobial resistance phenotype prediction (AMRPP) offers a computational alternative to traditional susceptibility testing.
- Acinetobacter baumannii is a critical priority pathogen with limited comprehensive computational resistance prediction studies.
- Existing AMRPP research often overlooks multi-antibiotic modeling and optimal strain representation.
Purpose of the Study:
- To introduce a multi-antibiotic AMRPP framework for Acinetobacter baumannii using gene presence/absence (GPA) profiles.
- To evaluate the efficacy of classical machine learning (ML) versus deep learning (DL) architectures for AMRPP.
- To explore and compare feature engineering strategies for GPA data, including KEGG filtering, information content (IC), principal component analysis (PCA), and SHapley Additive exPlanations (SHAP).
Main Methods:
- Developed a multi-antibiotic AMRPP framework for Acinetobacter baumannii.
- Encoded bacterial strains using gene presence/absence (GPA) profiles.
- Applied four feature enhancement strategies: KEGG pathway filtering, information content (IC) selection, PCA, and SHAP.
- Benchmarked ML models (SVM, Random Forest, XGBoost) against a custom DL model (TripSimAcin-AMR).
Main Results:
- Data-driven representations (PCA, IC, SHAP) outperformed KEGG filtering, achieving 92.64-93.16% performance.
- TripSimAcin-AMR and XGBoost showed comparable accuracy, but TripSimAcin-AMR demonstrated higher specificity (85.17% vs. 78.77%).
- The optimal configuration (TripSimAcin-AMR with IC features) reached 94.2% accuracy, 94.2% AUC-ROC, 97.0% sensitivity, and 91.3% specificity.
Conclusions:
- Enriched GPA representations are crucial for robust antimicrobial resistance prediction.
- The developed framework and DL model show significant potential for accurate and specific AMRPP in Acinetobacter baumannii.
- High specificity in AMRPP is clinically vital to prevent misinformed antibiotic treatment decisions.
Related Concept Videos
01:25Clinical Significance of Antibiotic Resistance
Methicillin-resistant Staphylococcus aureus (MRSA) presents a critical public health threat, arising from its capacity to resist β-lactam antibiotics due to acquisition of the mecA gene within the staphylococcal cassette chromosome mec (SCCmec). This gene encodes penicillin-binding protein 2a (PBP2a), which impairs binding efficacy of methicillin and other β-lactams. MRSA has evolved into distinct clonal lineages impacting humans and animals alike, reinforcing its significance within the One...
01:25Mechanism of Antibiotic Resistance in MRSA
Antibiotic resistance in bacteria arises when microorganisms evolve the ability to withstand drugs designed to kill them or inhibit their growth, rendering once-effective treatments useless. This phenomenon, driven by genetic change and selection under antibiotic exposure, poses a profound threat to modern medicine. Mechanisms include drug-inactivating enzymes (e.g., β-lactamases), efflux pumps that eject antibiotics, mutations altering antibiotic targets, decreased drug uptake, and acquisition...
01:30Development of Antibiotic Resistance
Antibiotic resistance is a major public health concern that arises when bacteria evolve mechanisms to withstand the effects of antibiotic treatments. This resistance can be intrinsic, acquired through genetic mutations, or transferred between bacteria via horizontal gene transfer. The development of antibiotic resistance poses significant challenges in treating bacterial infections and necessitates ongoing research to develop new therapeutic strategies.Intrinsic resistance occurs when bacterial...