What is it about?
Sepsis, the body's harmful reaction to infection, can also suppress the bone marrow, lowering white cells, red cells or platelets. This 'myelosuppression' makes patients more vulnerable and increases the risk of death. We measured 13 amino acids in the blood of 390 patients admitted with sepsis to the internal medicine department of the Trieste University Hospital and used machine learning to see which features best identified those with myelosuppression. About 4 in 10 patients had it, and they had more inflammation, more bloodstream infections and higher 30-day mortality (20% vs 12%). Among 44 clinical and laboratory variables, glutamic acid was the strongest predictor: patients with low levels at admission (below about 134 micromol/L) were more likely to have bone marrow suppression and had lower survival.
Featured Image
Photo by Anirudh on Unsplash
Why is it important?
Sepsis lacks simple tools to identify, on arrival, the patients at highest risk. Our results suggest that plasma glutamic acid, measured at admission, could become an early biomarker to stratify risk and to explore how metabolism and the immune system interact in sepsis. Although causality cannot be proven, the findings open the way to longitudinal and interventional studies, including on nutritional and metabolic support.
Perspectives
Most septic patients in Italy are cared for in internal medicine wards, often very old and frail, and they are under-represented in research. I value this collaboration with the Clinica Medica in Trieste because it looks at sepsis from a metabolic angle that infectious diseases specialists rarely consider. Combining machine learning with classic biochemistry gave us an unexpected candidate biomarker.
Professor Stefano Di Bella
Universita degli Studi di Trieste
Read the Original
This page is a summary of: Plasma glutamic acid predicts myelosuppression and mortality in septic patients using machine learning, Scientific Reports, January 2026, Springer Science + Business Media,
DOI: 10.1038/s41598-025-34178-x.
You can read the full text:
Contributors
The following have contributed to this page
