Unravelling lumbar disc herniation severity beyond MRI : integrated transcriptomic and metabolomic analyses highlight glycerophospholipid metabolism and inform a machine-learning diagnostic model: a pilot study.

Abstract

Aims: While MRI serves as a tool for assessing the severity of lumbar disc herniation (LDH), it has been observed that imaging diagnoses do not always align with clinical symptoms in nearly half of patients. The absence of dependable prognostic biomarkers impedes the early and accurate diagnosis of LDH, which is critical for the development of further treatment approaches. Thus, the aim of this study was to elucidate the molecular mechanisms that determine pain and LDH severity.

Methods: We conducted a pilot study with 55 patients, employing transcriptomic and metabolomic analyses on blood samples to identify potential biomarkers. A gene-metabolite interaction approach helped in identifying the pivotal pathway linked to disease severity. Moreover, a machine-learning model was designed to differentiate between patients based on the intensity of pain.

Results: Cholinergic-related glycerophospholipid metabolism emerged as the predominant enriched pathway in the severe symptom group via gene-metabolite interaction network analysis. Among various models, the gradient boosting machines (GBM) model stood out, achieving a commendable area under the curve (AUC) of 0.875 in distinguishing between the severe and mild symptom groups using combined RNA and metabolomics data.

Conclusion: Integrated molecular profiling of blood biomarkers has highlighted a novel determining pathway for LDH severity. This machine-learning approach can serve as a valuable predictive tool when MRI findings are inconclusive. Future research will focus on validating these biomarkers and exploring their potential for personalized medicine approaches.