By Raffaele Pereno – PhD, MBA
BASEL, Switzerland – At BioTechX Europe 2026, CheckOrphan met Dr. Mamadou Bamba, a researcher at De Montfort University in Leicester, UK, whose work explores how artificial intelligence (AI) could help identify biomarkers for rare diseases. His research into Niemann–Pick disease type C1 (NPC1) highlights adipate, a molecule linked to fatty-acid metabolism, as a potential blood biomarker.
The Challenge of Early Diagnosis
NPC1 is a rare inherited disorder in which cholesterol and other fats accumulate abnormally inside cells. Symptoms vary widely, from enlarged liver or spleen in young children to difficulties with movement, coordination, learning and memory. This variability can delay diagnosis, sometimes for years.
“For patients and families affected by NPC1, the challenge is that early symptoms can resemble those of other conditions. Identifying measurable changes in blood could help us recognize the disease sooner and direct patients toward specialist assessment,” said Bamba.
Using AI to Discover New Biomarkers
Bamba analyzed metabolic profiles from blood samples of 59 NPC1 patients and 71 controls, alongside data from a mouse model of the disease. Machine-learning methods searched for metabolic patterns that distinguished affected individuals from controls.
The model reportedly distinguished patients from controls with approximately 83% accuracy in the analyzed dataset. An approach called explainable AI then helped identify the molecules that contributed most to the model’s predictions. Adipate emerged as a promising candidate.
“The value of explainable AI is that it does more than classify samples: it helps us identify the metabolic signals driving those predictions. Adipate stood out as a candidate worth investigating further, although we still need to understand its biological relationship with NPC1,” Bamba explained.
From Discovery to Clinical Application
Adipate is not yet a validated diagnostic test for NPC1. Larger, independent studies are needed to establish whether it can reliably distinguish patients from people without the disease, and whether its levels reflect disease severity or progression.
For rare diseases, where patient numbers are limited and early diagnosis is difficult, combining metabolomics with explainable AI could help researchers find overlooked biological clues and develop better tools for patients.
For more information, contact
Dr. Mamadou Bamba
De Montfort University, Leicester, United Kingdom