Serum Metabolomics for Multi-Disease Classification in T2D, IBD, and CRC

517.8 KB of processed metabolomics data from three public studies (ST003390, ST003312, ST000284) for Type 2 diabetes, inflammatory bowel disease, and colorectal cancer. The dataset, authored by Rabmit Das and last updated in April 2026, contains results from a standardized machine learning workflow applied to serum samples. It includes benchmarked classification results and metabolite-to-pathway mappings supporting proteome-metabolome network analysis.

Use Cases

• Benchmarking classification algorithms based on the eight models (e.g., logistic regression, XGBoost) described in the study.
• Identifying disease-linked metabolic signatures based on the glucose, lipid, amino acid, and nucleotide pathways mentioned.
• Developing interpretable diagnostic strategies based on the pathway-level markers derived from KEGG mapping.

Strengths

• Data is processed with a standardized workflow including random forest imputation, batch correction, and Pareto scaling.
• Classification performance is quantified with specific metrics, including near-perfect AUC for T2D and multiclass accuracy >0.9.
• Results are mapped to specific KEGG pathways (e.g., glucose metabolism in T2D) for biological interpretation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale model training.
• The 517.8 KB size indicates a limited scope, likely containing summary or processed results rather than raw spectral data.

Provenance

Source

Publicly available serum metabolomics datasets from figshare (ST003390, ST003312, ST000284).

Collection Method

Standardized workflow combining imputation, transformation, scaling, batch correction, and supervised machine learning.

Time Range

null

Freshness

Last updated 2026-04-23 15:06:20; freshness should be verified.

Geography

null