A recent study by CAMO-Net Uganda, based in the the Infectious Diseases Institute (IDI), Makerere University, has shown that resistance to antibiotics is rising in Uganda.
As part of our efforts to address this trend, the CAMO-Net Uganda national hub is developing machine learning (ML) models designed to address antimicrobial resistance (AMR) in the region. These models, led at IDI by Dr Ronald Galiwango, are designed to predict clinical outcomes in patients with drug-resistant infections and will improve patient management, support AMR surveillance in the region, and optimise resource allocation for AMR interventions.
Antimicrobial resistance is a growing global health crisis, rendering existing treatments ineffective and leading to increased mortality, prolonged hospital stays, and rising healthcare costs. If left unaddressed, AMR could cause 1.91 million direct deaths and contribute to 8.22 million deaths globally by 2050. AMR occurs when bacteria, viruses, fungi, and parasites develop resistance to antimicrobial drugs, largely due to the misuse and overuse of antibiotics in human, animal, and environmental settings.
Uganda faces significant challenges in tackling AMR. An IDI study, led by Dr Jonathan Mayito, CAMO-Net researcher and internal medicine physician, and conducted across nine Regional Referral Hospitals (RRHs), revealed rising resistance across all drug categories, including the Access, Watch, and Reserve (AWaRe) classification. One of the main contributors to AMR in Uganda is the widespread availability of over-the-counter antibiotics, leading to self-medication, incomplete treatment courses, and misuse for non-bacterial infections.
Additionally, delays in antimicrobial susceptibility testing (AST) force healthcare providers to rely on empirical treatment, increasing the misuse of broad-spectrum antibiotics and accelerating resistance. Weak data systems further hinder AMR surveillance and effective decision-making. Addressing these issues—reducing AST turnaround times, improving antibiotic stewardship, and strengthening data systems—is critical for optimising antimicrobial use, improving patient outcomes, and curbing resistance.
Machine learning and AMR
Machine learning (ML) is gaining traction as a promising approach to addressing AMR, driven by advancements in computational power, algorithm performance, and expanding clinical datasets. Unlike traditional methods, which are often slow, expensive, and labour-intensive, ML can rapidly process vast datasets—including genomic sequences, clinical records, and molecular structures—to generate accurate predictions and insights. ML models are dynamic and adaptive, making them well-suited for tackling the constantly evolving nature of AMR.
ML is a branch of artificial intelligence (AI) that enables computers to learn patterns from data and make predictions without explicit programming. There are two primary approaches:
1. Supervised learning, where the model is trained on labelled data with known outcomes.
2. Unsupervised learning, where the model detects patterns in unlabelled data, identifying clusters and relationships.
Applications of ML in addressing AMR
1. Drug discovery and design
Traditional antibiotic discovery is slow and costly, often taking over a decade and billions of dollars to develop new drugs. ML accelerates this process by screening vast chemical libraries and predicting antimicrobial compounds with desirable properties. Deep learning techniques, such as neural networks, enhance drug-target interaction modelling, aiding in the development of new antibiotics. ML also integrates with tools like AlphaFold and molecular docking to predict how drugs interact with bacterial proteins, facilitating more effective treatment options.
Combination therapy, which uses multiple drugs to combat infections, is a crucial strategy against AMR. Identifying effective drug combinations is challenging due to the vast number of possible interactions. ML simplifies this process by analysing millions of potential drug pairings and identifying synergistic effects, improving treatment efficacy and reducing resistance development.
2. Predicting resistance patterns
ML algorithms analyse genomic and epidemiological data to predict bacterial resistance patterns with high precision. By detecting genetic mutations linked to resistance, these models help researchers anticipate emerging resistance trends. This proactive approach enables public health authorities to guide antibiotic stewardship and implement effective infection control strategies. By forecasting resistance mechanisms, ML supports better decision-making to curb AMR at both national and global levels.
3. Optimising treatment strategies
Machine learning enhances clinical decision-making by analysing patient-specific data—such as medical history, microbiology results, and demographics—to recommend the most effective antibiotics and dosages. By predicting the likelihood of resistance and optimising treatment plans, ML-powered clinical decision support tools help reduce unnecessary antibiotic use, minimise side effects, and improve patient outcomes.
Challenges in Applying ML to AMR
Despite its potential, machine learning faces several challenges in AMR research:
- Model Interpretability: Many ML models operate as “black boxes,” making it difficult to understand how they generate predictions. Clinicians need transparency to trust and adopt these models.
- Data Availability and Quality: ML models require large, high-quality datasets, but healthcare data is often fragmented, inconsistent, and error-prone, limiting model accuracy.
- Bias and Generalisability: If training data does not represent the target population, the models may develop biases, reducing their real-world applicability
The CAMO-Net Uganda national hub is working to provide solutions to these problems. Machine learning offers transformative potential in our efforts to address AMR by accelerating drug discovery, predicting resistance patterns, and optimising treatment strategies.
Dr Ronald Galiwango, who leads the machine learning research at CAMO-Net Uganda said, “Machine learning has the potential to transform how we address antimicrobial resistance by providing faster, more precise insights into resistance patterns and treatment outcomes. By harnessing vast datasets, we can move beyond reactive approaches and develop proactive, data-driven strategies that optimise antibiotic use and improve patient care. Our work at CAMO-Net Uganda is focused on building ML models that not only predict resistance trends but also support clinicians in making more informed decisions—helping to slow the spread of resistance and safeguard the effectiveness of existing treatments.”
Ongoing investment in data infrastructure, workforce training, and policy integration is essential for realising machine learning’s full potential in the global effort to address AMR. By leveraging these capabilities, Uganda and other regions disproportionately affected by AMR can enhance its management, ensuring more effective antibiotic use and better patient outcomes.
