Daily Sepsis Research Analysis
Three studies advance sepsis precision medicine and frontline care: a prospective multicenter machine-learning study defines clinically actionable sepsis phenotypes aligned with the plasma proteome; a translational multi-omics investigation identifies urinary 3‑methylhistidine as a promising biomarker for sepsis‑associated acute kidney injury; and Ugandan cohort data highlight underrecognized rickettsial infections in febrile sepsis with diagnostic rRNA RT‑PCR performance and clear empirical tre
Summary
Three studies advance sepsis precision medicine and frontline care: a prospective multicenter machine-learning study defines clinically actionable sepsis phenotypes aligned with the plasma proteome; a translational multi-omics investigation identifies urinary 3‑methylhistidine as a promising biomarker for sepsis‑associated acute kidney injury; and Ugandan cohort data highlight underrecognized rickettsial infections in febrile sepsis with diagnostic rRNA RT‑PCR performance and clear empirical treatment implications.
Research Themes
- Data-driven sepsis phenotyping and proteomics
- Biomarkers for sepsis-associated acute kidney injury
- Underrecognized infectious etiologies and empiric therapy in sepsis
Selected Articles
1. Machine learning identifies clinical sepsis phenotypes that translate to the plasma proteome.
A prospective multicenter cohort (n=384) used machine learning to define three sepsis phenotypes with distinct organ failure patterns and mortality risk, mirrored by graded consumption of complement and coagulation proteins. A supervised classifier using seven routine variables enabled early phenotype assignment, creating a path toward phenotype-guided trials.
Impact: Provides a scalable clinical-omics framework to stratify sepsis, enabling hypothesis-driven, phenotype-specific therapies and more efficient trial design.
Clinical Implications: Early phenotype assignment using seven routine variables could aid triage, prognostication, and selection for phenotype-enriched trials. Proteomic signatures point to targetable pathways (complement/coagulation) for precision therapeutics.
Key Findings
- Three sepsis phenotypes were identified; cluster C had highest severity with liver failure and strongest mortality association.
- Plasma proteomics showed graded consumption of complement and coagulation factors with increasing severity across phenotypes.
- A supervised ML model classified phenotypes using seven widely available features (ALT, AST, BE, INR, BPsys, BPdia, aPTT).
Methodological Strengths
- Prospective multicenter cohort with harmonized sampling and mass spectrometry-based proteomics.
- Transparent ML approach linking clinical phenotypes to mechanistic proteomic pathways.
Limitations
- Moderate sample size and single healthcare geography may limit generalizability.
- No interventional validation; phenotype-specific treatment effects remain untested.
Future Directions: External validation across health systems, prospective phenotype-guided trials, and integration of proteomic signatures to select targeted therapies.
BACKGROUND: Sepsis therapy is still limited to treatment of the underlying infection and supportive measures. To date, various sepsis subtypes were proposed, but therapeutic options addressing the molecular changes of sepsis were not identified. With the aim of a future individualized therapy, we used machine learning (ML) to identify clinical phenotypes and their temporal development in a prospective, multicenter sepsis cohort and characterized them using plasma proteomics. METHODS: Routine clinical data and blood samples were collected from 384 patients. Sepsis phenotypes were identified based on clinical measurements and plasma samples from 301 patients were analyzed using mass spectrometry. The obtained data were evaluated in relation to the phenotypes, and supervised ML models were developed enabling prospective phenotype classification. RESULTS: Three sepsis phenotypes were identified. Cluster C was characterized by the highest disease severity and multi-organ failure with leading liver failure. Cluster B showed relevant organ failure, with renal damage being particularly prominent in comparison to cluster A. Time course analysis showed a strong association of cluster C with mortality, while patients in cluster B were likely to change the cluster until day 4. The plasma proteome reflected the clinical features of the phenotypes and showed gradual consumption of complement and coagulation factors with increasing sepsis severity. Supervised ML models allowed the assignment of patients based on only seven widely available features (alanine transaminase (ALT), aspartate transaminase (AST), base excess (BE), international normalized ratio of thrombin time (INR), diastolic arterial blood pressure, systolic arterial blood pressure (BPdia, BPsys) and activated partial thromboplastin time (aPTT)). CONCLUSIONS: The identified clinical phenotypes reflected varying degrees of sepsis severity and were mirrored in the plasma proteome. Proteomic profiling offered novel insights into the molecular mechanisms underlying sepsis and enabled a deeper characterization of the identified phenotypes.
2. Urinary 3-methylhistidine as a potential biomarker for sepsis-associated acute kidney injury: multidimensional metabolomics analysis in mice and human.
Integrating mouse RT‑GFR models, untargeted and spatiotemporal renal metabolomics, and a human urine cohort (n=95), the study identifies urinary 3‑methylhistidine as an SA‑AKI biomarker with AUC 0.86 and a combined clinical model AUC 0.89. Spatial analyses localize 3‑MH increases to collecting ducts, supporting biological plausibility.
Impact: Delivers a mechanistically anchored, noninvasive biomarker with immediate diagnostic potential for SA‑AKI—an unmet clinical need in sepsis care.
Clinical Implications: Urinary 3‑MH could support early SA‑AKI screening and risk stratification, enabling timely nephroprotective strategies and enrollment into AKI‑focused trials.
Key Findings
- Urinary 3‑methylhistidine identified as a key metabolite with diagnostic AUC 0.86 (95% CI 0.77–0.95) for SA‑AKI; combined model AUC 0.89.
- Renal spatiotemporal metabolomics localized increased 3‑MH distribution to collecting ducts in SA‑AKI.
- Multi-omics pipeline (mouse RT‑GFR model, tissue/urine metabolomics, human cohort) yielded a biologically plausible, translatable biomarker.
Methodological Strengths
- Integrated multi-omics with both animal models and human validation, including spatial metabolomics.
- Use of RT‑GFR for precise AKI phenotyping and robust machine-learning feature selection.
Limitations
- Human cohort size was modest and from a single setting; external validation is needed.
- Cross-sectional sampling limits assessment of temporal kinetics and clinical cutoffs.
Future Directions: Prospective multicenter validation with serial sampling, assay standardization, and clinical impact studies on early intervention guided by 3‑MH.
BACKGROUND: Sepsis-associated acute kidney injury (SA-AKI) is strongly associated with increased mortality in critical patients. The early detection of SA-AKI is crucial for clinical intervention. This study aims to integrate multiple metabolomics data related to SA-AKI to identify and validate novel metabolic markers. METHODS: Real-time glomerular filtration rate (RT-GFR) measurement was adopted to establish SA-AKI mice. Untargeted metabolomics sequencing was performed on SA-AKI mice renal tissue (Control-LPS-8 h-LPS-24 h, N = 4) and urine samples (Control group vs. LPS-24 h group, N = 6). Time series analysis and random forest algorithm were employed to identify key metabolic molecule. Subsequently, renal spatiotemporal metabolomics was used to explore the specific distribution of key molecule. Eventually, a clinical cohort (20 healthy volunteers vs. 30 sepsis patients vs. 45 SA-AKI patients) urine quantitative metabolomic analysis was carried out to validate it as a biomarker and construct a diagnostic model via logistic regression (LR). RESULTS: Forty-two key renal metabolites and top fifty urinary metabolites were determined through multidimensional metabolomics study of SA-AKI mice. Urinary 3-Methylhistidine (3-MH) was charactered as a potential biomarker. The distribution of 3-MH increased in collecting ducts through renal spatiotemporal metabolomics sequencing. Then, we recruited 95 urine samples to validate its diagnostic performance (AUC = 0.86, 95% CI 0.77-0.95) and its role as an independent predictive factor for SA-AKI (OR = 0.21, 95% CI: 0.05-0.84, p < 0.05). Ultimately, a diagnostic model combined urinary 3-MH with clinical variables was constructed to identify SA-AKI (AUC = 0.89, 95% CI 0.74-1.00). CONCLUSIONS: We proposed that urinary 3-Methylhistidine has potential diagnostic value for SA-AKI screening. Future studies will focus on its performance in other clinical populations to comprehensively evaluate its diagnostic role.
3. Rickettsioses as Underrecognized Cause of Hospitalization for Febrile Illness, Uganda.
Analyzing archived samples from Ugandan sepsis and acute febrile illness cohorts (n=329), 10% had rickettsioses. Serum rRNA RT‑PCR showed 75% sensitivity and 91% specificity; thrombocytopenia was strongly associated. Findings support including doxycycline empirically for nonmalarial febrile illness and deploying rRNA RT‑PCR diagnostics.
Impact: Directly informs empiric antibiotic choices and diagnostic strategies for febrile sepsis in sub‑Saharan Africa, where rickettsioses are underrecognized.
Clinical Implications: In regions with high nonmalarial febrile illness, empiric doxycycline should be considered. Implementing rRNA RT‑PCR can improve early pathogen-directed therapy and reduce mortality.
Key Findings
- Rickettsial infections accounted for 10% (33/329) of sepsis/AFI presentations in Uganda.
- Serum rRNA RT‑PCR had 75.0% sensitivity and 91.2% specificity against reference methods.
- Thrombocytopenia was significantly associated with rickettsioses (adjusted OR 3.7; p=0.003), and no patients received tetracycline at admission, supporting empiric doxycycline.
Methodological Strengths
- Dual diagnostic approach using immunofluorescence assay and clinically validated rRNA RT‑PCR on cohort samples.
- Real-world cohorts spanning sepsis and acute febrile illness increase applicability to frontline care.
Limitations
- Use of archived samples limits control over timing relative to symptom onset.
- Regional generalizability may vary; lack of interventional antibiotic outcome data.
Future Directions: Prospective implementation studies of rRNA RT‑PCR in emergency workflows and randomized evaluations of empiric doxycycline strategies in nonmalarial febrile illness.
The complexity of rickettsial serodiagnostics during acute illness has limited clinical characterization in Africa. We used archived samples from sepsis (n = 259) and acute febrile illness (n = 70) cohorts in Uganda to identify spotted fever and typhus group rickettsiae by using immunofluorescence assay and clinically validated rRNA reverse transcription PCR (RT-PCR). Among 329 participants, 10.0% had rickettsial infections (n = 33; n = 20 identified with immunofluorescence assay and n = 13 by RT-PCR). Serum rRNA RT-PCR was 75.0% (95% CI 42.8-94.5%) sensitive and 91.2% (95% CI 85.8-95.1%) specific. Thrombocytopenia was more common among patients with rickettsial infections than with other nonmalarial infections (adjusted odds ratio 3.7; p = 0.003). No participants were on a tetracycline antimicrobial drug at admission. rRNA RT-PCR is a promising diagnostic strategy for identifying acute rickettsial infections. Doxycycline should be included in empiric antimicrobial drug regimens for nonmalarial febrile illness in this region.