Serum Metabolomics for Intrinsic and Idiosyncratic Drug-Induced Liver Injury Biomarkers

A metabolomic study of 44 drug-induced liver injury (DILI) cases, comprising 17 intrinsic and 27 idiosyncratic types, conducted by Xianni Wei and last updated in March 2026. The dataset was generated using high-performance chemical isotope labeling liquid chromatography–mass spectrometry (HP-CIL LC-MS) to profile serum samples. It includes four identified differential metabolites and results from machine learning models used for diagnostic biomarker discovery.

Use Cases

• Training machine learning classifiers to distinguish intrinsic from idiosyncratic DILI based on serum metabolite profiles.
• Validating potential biomarkers like Alanyl-Glycine and N2-Acetyl-L-Cystathionine for clinical diagnostic applications.
• Analyzing metabolic pathway disruptions, particularly in amino acid metabolism pathways such as tryptophan and tyrosine metabolism.
• Benchmarking the performance of different machine learning models (e.g., ROC analysis, multivariate regression) on a small, focused clinical dataset.

Strengths

• Includes results from multiple machine learning models, with reported AUC values greater than 0.8 and up to 0.983 in internal validation.
• Identifies four specific differential metabolites (e.g., Alanyl-Glycine) with associated confidence levels (level 1, level 2a).
• Case classification follows established diagnostic criteria from EASL clinical practice guidelines and a NEJM review.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale model training.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is small in scale (288.4 KB), indicating a limited sample size of 44 cases.

Provenance

Source

figshare, author Xianni Wei.

Collection Method

Serum samples analyzed via HP-CIL LC-MS; cases classified per clinical guidelines.

Freshness

Last updated 2026-03-18 07:36:42; freshness should be verified.