Skip to contents

parkinsonsMetagenomicData

Package Overview

This package provides access to uniformly processed gut microbiome data from Parkinson’s Disease studies. The gut-brain axis has emerged as an important factor in Parkinson’s Disease, with multiple studies identifying differences in the gut microbiome composition between individuals with Parkinson’s and healthy controls.

What data is available?

This package provides comprehensive microbiome profiling data from fecal samples, including:

  • Microbial composition: Which bacteria, viruses, and other microorganisms are present and their relative abundances
  • Bacterial strain identification: Fine-resolution tracking of specific bacterial strains
  • Functional profiles: What genes and metabolic pathways are encoded by the gut microbiome
  • Quality metrics: Technical information about sequencing depth and data quality

All data have been uniformly processed through the curatedMetagenomicsNextflow pipeline using standardized bioinformatics tools (MetaPhlAn, StrainPhlAn, HUMAnN, FastQC, KneadData) to ensure consistency across studies.

For additional analysis utilities including taxonomic handling and statistical tests, see biobakeryUtils.

Available Data

The ASAP-MAC initiative has collected a number of Parkinson’s Disease-focused studies for metadata curation, uniform processing, and meta-analysis. The ongoing process of the collection of these datasets can be followed in the parkinsons_data_search repository. The majority of the studies listed in the table parkinson_shotgun_datasets.tsv are available for retrieval with this package.

To browse the available data, load the sampleMetadata object included in this package. To see which data types are available in the remote repositories, use get_repo_info() and get_hf_parquet_urls(). See the vignettes for examples.

Sample Metadata

Metadata for all samples is available through the sampleMetadata data frame, which includes curated clinical and demographic information (identifiers, age, sex, disease status, study design) as well as uncurated features (prefixed with “uncurated_”). The curation process can be seen in more detail in the parkinsonsManualCuration repository. For complete variable definitions, allowed values, and column descriptions, see the Data Codebook vignette.

Data Types

Data are organized into several categories based on biological content:

Taxonomic Composition (MetaPhlAn outputs) - relative_abundance: Relative abundance of bacterial, archaeal, viral, and eukaryotic species - viral_clusters: Viral community composition - marker_abundance, marker_presence: Species-specific genetic markers for taxonomic identification

Strain-Level Profiling (StrainPhlAn outputs) - strainphlan_markers: Strain-specific genetic markers for tracking bacterial variants

Functional Profiling (HUMAnN outputs) - genefamilies_*: Abundance of gene families (groups of related genes) - pathabundance_*: Abundance of metabolic pathways (e.g., carbohydrate metabolism, amino acid synthesis) - pathcoverage_*: Coverage/completeness of metabolic pathways

Normalization options: Raw counts, relative abundance (relab), or copies per million (cpm); some data are stratified by contributing species, others are community totals (unstratified).

Quality Control - fastqc: Sequencing quality metrics (read quality, GC content, adapter contamination) - kneaddata_log: Pre-processing statistics (human DNA removal, quality filtering)

Data Hosting

While the sample metadata are available within this package, the various output files are hosted remotely due to their size and number. There are therefore two options for data retrieval.

Google Cloud Storage (Developer Access)

The raw pipeline outputs are stored in the Google Cloud Bucket gs://metagenomics-mac, which requires credentials for access. This is primarily for developers and maintainers who need to access individual sample files directly from the pipeline. Most users should use the Hugging Face repository instead (see below).

For developers: Once authenticated, data are organized as individual files for each sample and output type, and can be downloaded with requester-pays GCP egress fees. The parquet_generation repository contains scripts for transforming these raw outputs into the consolidated parquet files.

For most users, the data have been combined into parquet files (see this process in the parquet_generation repository) and hosted publicly on Hugging Face in the metagenomics_mac repo. Smaller example files featuring data from 10 samples each can be found at metagenomics_mac_examples.

These files can be easily accessed through the DuckDB R client. The package provides convenient wrapper functions documented in the vignettes:

Linked Repositories

Here are all of the repositories involved in making this data available: