Sepsis Prognostic Prediction Panel Based on Serum Amino Acid Profiles

A dataset of serum amino acid profiles from 232 participants, including 60 healthy controls and 172 sepsis patients, used to develop a prognostic prediction model. The data was generated using quantitative mass spectrometry and analyzed with multiple machine learning methods by author Yue Liu. The dataset was last updated on June 1, 2026.

Use Cases

• Training prognostic prediction models for sepsis based on serum amino acid levels.
• Identifying candidate biomarkers for sepsis diagnosis based on amino acid profile alterations.
• Comparing the performance of different machine learning models (e.g., Deephit, XGBoost) for survival analysis in clinical data.
• Investigating the relationship between specific amino acids (e.g., glutamine, glycine) and clinical outcomes in sepsis.

Strengths

• Includes data from 232 participants (60 healthy controls, 172 sepsis patients) with a defined external test set of 130 patients.
• The prognostic model is based on five specific amino acid features (glutamine, glycine, lysine, pyroglutamic acid, proline) identified through feature importance ranking.
• Data was generated using quantitative (liquid chromatography) mass spectrometry, a standard analytical technique.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the primary data tables is unknown, which may limit suitability assessment.
• The dataset is relatively small at 4.1 MB, which may constrain model complexity or generalizability.

Provenance

Source

figshare, author Yue Liu.

Collection Method

Quantitative mass spectrometry analysis of serum samples from patients and healthy controls.

Freshness

Last updated 2026-06-01 05:22:31; freshness should be verified.