In Parkinson's disease (PD), gastrointestinal features are common and often precede the motor signs. Braak and colleagues proposed that PD may start in the gut, triggered by a pathogen, and spread to the brain. Numerous studies have examined the gut microbiome in PD, all found it to be altered, but found inconsistent results on associated microorganisms. Studies to date have been small (N=20 to 306) and are difficult to compare or combine due to varied methodology. We conducted a microbiome-wide association study (MWAS) with two large datasets for internal replication (N=333 and 507). We used uniform methodology when possible, interrogated confounders, and applied two statistical tests for concordance, followed by correlation network analysis to infer interactions. Fifteen genera were associated with PD at a microbiome-wide significance level, in both datasets, with both methods, with or without covariate adjustment. The associations were not independent, rather represented 3 clusters of co-occurring microorganisms. Cluster 1 was composed of opportunistic pathogens; all were elevated in PD. Cluster 2 were short-chain-fatty-acid producing bacteria; all were reduced in PD.Cluster 3 were carbohydrate-metabolizing probiotics; elevated in PD. Depletion of antiinflammatory short-chain-fatty-acid producing bacteria and elevated levels of probiotics are confirmatory. Overabundance of opportunistic pathogens is a novel finding and their identity provides a lead to experimentally test their role in PD. animal models. [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] Every study that has compared the global composition of the gut microbiome in PD vs. controls found it to be significantly altered; in contrast, attempts to identify PD-associated microorganisms have produced inconsistent results. 31,32 Low reproducibility has been attributed to small sample sizes (missing true associations due to low power), relaxed statistical thresholds (inflating false positive results), and publishing without a replication dataset (required for genomic studies). Differences in methods of DNA extraction, sequencing, bioinformatics and statistics can all contribute to inter-study variations. The choice of taxonomic resolution for analysis (PD has been tested at all levels from phylum to species) and the inconsistent taxonomic assignments and nomenclature used in various reference databases add to the confusion when comparing results. Last but not least, is confounding by heterogeneity in the populations that were studied: PD is heterogenous and so is the microbiome. PD subtypes cannot be readily identified thus patient populations are inevitably varied. A myriad of factors can affect the microbiome ranging from diet, health and medication to cultural habits, life-styles, race and geography. 33,34 Identifying microorganisms involved in the dysbiosis of the microbiome is essential for understanding their role in disease. We conducted a hypothesis-free microbiome-wide association study (MWAS) modeled after and usi...