Machine learning-based meta-analysis reveals gut microbiome alterations associated with Parkinson’s disease
Stefano Romano, Jakob Wirbel, Rebecca Ansorge, Christian Schudoma, Quinten R. Ducarmon, Arjan Narbad, Georg Zeller
Abstract
There is strong interest in using the gut microbiome for Parkinson's disease (PD) diagnosis and treatment. However, a consensus on PD-associated microbiome features and a multi-study assessment of their diagnostic value is lacking. Here, we present a machine learning meta-analysis of PD microbiome studies of unprecedented scale (4489 samples). Within most studies, microbiome-based machine learning models accurately classify PD patients (average AUC 71.9%). However, these models are study-specific and do not generalise well across other studies (average AUC 61%). Training models on multiple datasets improves their generalizability (average LOSO AUC 68%) and disease specificity as assessed against microbiomes from other neurodegenerative diseases. Moreover, meta-analysis of shotgun metagenomes delineates PD-associated microbial pathways potentially contributing to gut health deterioration and favouring the translocation of pathogenic molecules along the gut-brain axis. Strikingly, microbial pathways for solvent and pesticide biotransformation are enriched in PD. These results align with epidemiological evidence that exposure to these molecules increases PD risk and raise the question of whether gut microbes modulate their toxicity. Here, we offer the most comprehensive overview to date about the PD gut microbiome and provide future reference for its diagnostic and functional potential.