Integrative Analysis of Neuroimaging and Microbiome Data Predicts Cognitive Decline in Parkinson’s Disease

  1. Büşranur Delice
  2. Özkan Ufuk Nalbantoğlu
  3. Süleyman Yıldırım

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by motor and non-motor symptoms, including cognitive impairment (CI) ranging from mild cognitive impairment (MCI) to Parkinson’s disease dementia (PDD). Growing evidence supports the gut-brain axis as playing an essential role in the pathophysiology of PD, suggesting promising applications for combining advanced neuroimaging techniques with microbiome profiling to accelerate biomarker discovery and therapeutic innovation. This study combines resting-state functional magnetic resonance imaging (rs-fMRI) and 16S rRNA sequencing of stool and saliva to identify biomarkers predictive of CI in PD.

A stepwise feature selection pipeline, incorporating ANOVA, random forest ranking, and partial correlation analysis, was applied to extract biologically meaningful features from rs-fMRI connectivity matrices and microbial taxa. Independent and joint machine learning models, including Random Forest, support vector machine, XGBoost, and logistic regression, were evaluated for their predictive performance. The joint model, integrating neuroimaging and microbiome features, outperformed modality-specific models in classifying HC, MCI, and PDD stages, achieving an accuracy of 88.9% and AUC of 97.2% with Random Forest. Key fMRI features involved the salience and default mode networks, while microbial biomarkers included taxa such as Faecalibacterium, Veillonella, and Streptococcus. Correlations between microbial taxa and fMRI features suggest potential gut-brain interactions influencing CI. For example, Faecalibacterium abundance was positively associated with connectivity in the salience network, while Veillonella showed links to executive function networks. These findings support the synergistic value of integrating multi-omics data for uncovering mechanisms underlying CI in PD.

This study demonstrates the utility of combining neuroimaging and microbiome data to enhance predictive performance and biological insight. The identified biomarkers may serve as a foundation for developing microbiome-targeted interventions and neuroimaging-guided strategies for managing cognitive decline in PD. Future work should focus on larger, longitudinal datasets and explainable AI approaches to further refine this integrative methodology.

Article activity feed

  1. PREreview

    This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/15533606.

    Brief summary of the study 

    • The study combines the typical method for testing Parkinson's diseases (PD) i.e. neuroimaging, with 16S rRNA sequencing of stool and saliva to identify biomarkers that can predict cognitive decline among patients. 

    • The study uses advanced machine learning (ML) techniques to identify patterns in brain structure and specific gut bacteria associated with faster cognitive decline in PD patients, with better predictive accuracy using combined data than either neuroimaging or microbiome data alone.

    Major comments

    Plus points

    • The study clearly mentions that ethics approval and informed consent was obtained

    • Cohort establishment and participants recruitment is clearly explained

    • Figures are well made, they are of high quality and text is readable 

    • The study utilizes a novel approach combining brain imaging and microbiome data to identify biomarkers that can be predictive of Parkinson's diseases progression

    Things to improve

    • Information on microbiome data analysis is scant. Description of saliva and stool collection procedure, DNA extraction and handling, DNA sequencing information and initial DNA data analysis including quality control is missing. 

    Minor comments 

    • Suggestions for future work could go in the discussion section, not in abstract, unless the scientist plan to conduct the studies 

    Comments on reporting 

    • Statistical analyses are well described 

    • Information on availability of genomic data is provided, but not for brain imaging data

    Conflicts of interest of reviewers

    • None declared 

    Competing interests

    The authors declare that they have no competing interests.

  2. Version published to 10.1101/2025.03.28.645180 on bioRxiv