Nigeria Researchers Urge Locally Trained AI Models for Health Data

Health researchers meeting in Abuja have called on African countries to build and train their own artificial intelligence models rather than rely on systems trained on external datasets, warning that imported models misread African genetic and pathogen diversity. Developing locally trained AI models is critical to ensuring clinical tools accurately reflect the continent’s distinct biological variations.

Diagnostic Discrepancies and the Need for Locally Trained AI Models

Oladipo Kolawole, associate professor of infectious diseases and founder of the Helix Biogen Institute, said global AI models predicted ciprofloxacin resistance with about 75 per cent accuracy in England but only about 50 per cent accuracy when applied to African data. “One size does not fit all; we must own our data and our AI tools,” he said. Without locally trained AI models, diagnostic accuracy in local clinical settings remains compromised.

Symposium Proceedings and Metadata Governance

He was speaking at the fourth annual scientific symposium of the Institute of Human Virology Nigeria, the News Agency of Nigeria reported. Alash’le Abimiku, who chaired the session, stressed the importance of comprehensive metadata collection alongside model development.

Whole Genome Sequencing, Machine Learning, and Biosecurity

Participants discussed combining whole genome sequencing with machine learning to predict disease trends and antimicrobial resistance, and the biosecurity measures research institutions are putting in place as genomic datasets grow. The integration of locally trained AI models with genomic surveillance promises to enhance national outbreak readiness.

Patient Safety and Clinical Impact

A model that is half as accurate on African data will misdirect antibiotic choices, making local training a patient safety question. Adopting locally trained AI models ensures that empirical treatment protocols align with real-world pathogen profiles.

Frequently Asked Questions

Why do foreign AI models perform poorly on African health data? Global AI models are primarily trained on Western genomic and clinical datasets, causing them to miss unique African genetic diversity, distinct pathogen strains, and regional antimicrobial resistance patterns.

How does whole genome sequencing improve AI predictive accuracy? Combining whole genome sequencing with machine learning allows AI tools to identify specific genetic mutations linked to drug resistance, enabling precise predictions of disease trends and effective treatment options.

 

Source: News Agency of Nigeria, 31 August 2026.