Objective: To investigate fecal microbiome changes in Parkinson’s disease (PD) vs. controls by meta-analysis of 16S rRNA gene sequences from different studies.

Background: The role of the gut microbiome in PD has garnered significant attention over the past five years, with at least 25 case-control fecal microbiome studies published to date. However, studies were often limited by small sample size, and differences in study populations, designs, and biostatistical analyses have made generalization of the findings difficult.

Method: A PubMed search was conducted to identify papers published from January 1, 2000 to May 1, 2020. After filtering of ineligible studies, raw 16S rRNA gene sequences of included studies were downloaded and quality-filtered using standardized parameters with DADA2. Denoised gene sequences were then analyzed using a phylogenetic placement approach. This method of pooling and reanalysis allows the harmonization of sequences generated at variable length with different primers, thus reducing the disparities in the analytical approach among studies. All statistical analyses were conducted in R. Differences in Aitchison distance-based beta diversity of microbial communities were tested by ADONIS using the vegan package. Differentially abundant bacterial taxa (adjusted p-value < 0.01) were identified based on negative binomial generalized linear model, using the DESeq2 package. The performance of the microbial composition in classifying controls and PD was evaluated using machine-learning models, including random forest, neural network, SVM and naïve Bayes.

Results: Ten studies (from nine countries) with a total of 969 PD patients and 734 non-PD controls were included in the meta-analysis. There were significant differences in fecal microbiome composition between cohorts from different studies. In the pooled analysis, PD status had a significant effect on fecal microbiome composition but explained only 0.3% of variances in the cohort. Genera with the highest fold changes in PD vs. controls included Acidaminococcus, Veillonella and Akkermansia (all increased), and Butyrivibrio (decreased). Neural network, built using differentially abundant genera, achieved an area under the curve of 0.69 with an accuracy of 0.65.

Conclusion: Fecal microbiome composition in PD significantly differs from non-PD controls, but the effect size is small, which may be due the influence of various confounders.

« Back to MDS Virtual Congress 2021

MDS Abstracts - https://www.mdsabstracts.org/abstract/piecing-together-the-gut-microbiome-in-parkinsons-disease-a-meta-analysis/