Background and Aims: Porto-sinusoidal vascular disorder (PSVD) is a rare and poorly understood vascular liver disease characterized by specific histological alterations in the absence of cirrhosis. This study aimed to evaluate differences in the metabolomic signature between patients with PSVD, cirrhosis, and healthy volunteers (HV) to guide non-invasive diagnosis and to gain further pathophysiological insights.
Method: We analyzed serum samples from 20 patients with histologically verified PSVD, cirrhosis, and HV. Metabolites were measured with liquid chromatography with mass spectrometry. Differential metabolite abundance was quantified with the LIMMA algorithm. Artificial neural network (ANN) and polynomial support vector machine (SVM) models were developed to classify PSVD vs. cirrhosis or HV metabolomic profiles.
Results: 281 metabolites were included in downstream analysis. Clustering separated PSVD metabolomes from HV, while a subset of PSVD patients (n=5, 25%) showed overlaps with cirrhosis. Differential testing revealed that aspartic acid, cholesterol, isovaleric acid, and hydroxyglutaric acid were linked to PSVD, whereas taurocholic and isovaleric acids to cirrhosis. Alterations in the pyrimidine, glycine, serine, and threonine pathways were exclusively associated with PSVD. Selected metabolic signatures optimized by a machine learning approach reliably distinguished PSVD from HV or cirrhosis utilizing 4 to 6 metabolites and were even further improved by including liver stiffness measurements.
Conclusion: High-throughput metabolomics allowed the identification of metabolic profiles in PSVD and cirrhosis. We observed alterations in the glycine, serine and threonine pathway potentially involved in the pathogenesis of PSVD. The identified metabolic signatures are candidates for evaluation in larger patient cohorts.