Meta2DB: curated shotgun metagenomic feature sets and metadata for health state prediction
TL;DR
Meta2DB is a newly developed, highly curated database of microbiome data and metadata designed to help train machine learning models for predicting human health conditions.
Problem / question
Publicly available metagenomic data and its associated metadata are often not standardized across different studies, which creates a barrier for researchers trying to train machine learning models to predict health outcomes.
Methods
The team utilized high-performance computing to standardize fifty terabytes of public raw sequencing data. They classified these sequences using a custom reference index containing all life kingdoms from an early 2023 snapshot of the NCBI Nucleotide database, and they combined manual and automated techniques to organize the associated metadata.
Key findings
The authors successfully built Meta2DB, a resource containing standardized microbiome taxonomy counts and metadata covering nearly fourteen thousand samples from over eighty studies, representing twenty-three diseases and thirty-four geographic regions.
Why it matters
Creating a massive and standardized dataset allows researchers to more easily build and train machine learning algorithms focused on diagnosing or predicting human health conditions based on microbiome profiles.
Limitations
The provided text does not mention any limitations of the database or the methodology.
Takeaway
Meta2DB offers a massive, standardized collection of microbiome data and metadata that is specifically optimized to support machine learning research in human health prediction.