Daily Sepsis Research Analysis
Three studies advance sepsis care across therapy, prediction, and diagnostics: a phage-derived depolymerase synergized with polymyxin B to rescue murine pandrug-resistant Acinetobacter baumannii bacteremia; an interpretable multi-cohort machine learning model accurately predicted persistent sepsis-associated AKI and outperformed urinary CCL14; and a prospective validation in Zimbabwe showed the BCID2 panel achieved high specificity and actionable resistance detection in a low-resource setting.
Summary
Three studies advance sepsis care across therapy, prediction, and diagnostics: a phage-derived depolymerase synergized with polymyxin B to rescue murine pandrug-resistant Acinetobacter baumannii bacteremia; an interpretable multi-cohort machine learning model accurately predicted persistent sepsis-associated AKI and outperformed urinary CCL14; and a prospective validation in Zimbabwe showed the BCID2 panel achieved high specificity and actionable resistance detection in a low-resource setting.
Research Themes
- Adjunctive enzymatic therapy against multidrug-resistant sepsis pathogens
- Explainable AI for early prediction of persistent sepsis-associated AKI
- Rapid molecular diagnostics implementation in low-resource settings
Selected Articles
1. Depolymerase as a potent adjunct to polymyxin for targeting KL160 pandrug-resistant Acinetobacter baumannii in a murine bacteremia model.
A KL160-specific phage-derived depolymerase (DPO-HL) synergized with polymyxin B, reducing polymyxin MIC 16-fold and achieving 100% survival in a murine pandrug-resistant A. baumannii bacteremia model while lowering endotoxin levels. DPO-HL was plasma-stable, enhanced plasma bactericidal activity, eradicated mature biofilms, and showed acceptable safety in vitro and in vivo.
Impact: This is among the first demonstrations in a mammalian sepsis model that a capsular depolymerase can rescue lethal pandrug-resistant A. baumannii bacteremia via synergy with a last-line antibiotic. It opens a translational path for enzyme–antibiotic combinations against critical AMR pathogens.
Clinical Implications: If validated in humans, depolymerase–polymyxin combinations could reduce doses, toxicity, and resistance emergence against A. baumannii sepsis. Capsular typing (e.g., KL160) may guide adjuvant selection.
Key Findings
- DPO-HL was stable in human plasma and enhanced plasma bactericidal activity.
- Synergy with polymyxin B reduced polymyxin MIC by 16-fold and eradicated mature biofilms.
- Combination therapy (1.45 mg/kg DPO-HL + 0.5 mg/kg polymyxin B) achieved 100% survival and reduced endotoxin; DPO-HL monotherapy rescued 30%.
Methodological Strengths
- Comprehensive in vitro and in vivo evaluation including biofilm assays, plasma interaction, safety, and murine survival.
- Clear pharmacodynamic synergy quantified by MIC reduction and survival endpoints.
Limitations
- Preclinical murine model; human efficacy and immunogenicity are unknown.
- Targeted a single capsular type (KL160), limiting immediate generalizability across A. baumannii strains.
Future Directions: Evaluate immunogenicity, pharmacokinetics, and efficacy across diverse capsular types; design first-in-human safety studies and optimize combination dosing strategies.
2. Explainable Machine Learning Model for Predicting Persistent Sepsis-Associated Acute Kidney Injury: Development and Validation Study.
Using 46,097 sepsis patients across retrospective and prospective cohorts, a 12-variable GBM model predicted persistent SA-AKI with AUCs of 0.87–0.98 across validations and outperformed urinary CCL14 in a prospective cohort. The model is explainable (SHAP) and deployed as a web tool for bedside risk stratification.
Impact: Provides an interpretable, externally validated tool that can be integrated into ICU workflows to triage sepsis patients at high risk of persistent AKI, potentially informing early nephroprotective strategies.
Clinical Implications: Supports early nephrology consultation, conservative nephrotoxin use, and fluid/diuretic stewardship in patients flagged high risk for persistent SA-AKI, beyond reliance on biomarkers alone.
Key Findings
- Final GBM model with 12 routine variables achieved AUC 0.870 (internal), 0.891 (MIMIC-III subset), 0.932 (eICU), and 0.983 (single-center external retrospective).
- In a prospective cohort, the GBM (AUC 0.852) outperformed urinary CCL14 (AUC 0.821) for predicting persistent SA-AKI.
- Model explainability via SHAP highlighted AKI stage, ΔCreatinine, urine output, and diuretic dose as top contributors; a web tool was released.
Methodological Strengths
- Multi-cohort design with internal, multiple external, and prospective validation.
- Explainability (SHAP) and deployment as an accessible clinical web tool.
Limitations
- Observational data may harbor residual confounding and site-specific practice biases.
- Generalizability to non-participating healthcare systems and low-resource settings requires further testing.
Future Directions: Prospective impact studies to test whether model-guided care reduces persistent SA-AKI and dialysis; adaptation and validation in low-resource ICUs.
3. Rapid bacterial identification and resistance detection using a low complexity molecular diagnostic platform in Zimbabwe.
In a 5-month prospective validation including 377 analyzable positive blood cultures, BCID2 achieved >95% specificity and organism-dependent sensitivity (50%–100%) in Zimbabwe. It identified widespread CTX-M (74.5% of Enterobacterales) and detected some NDM/VIM carbapenemases, with good usability (SUS 79.5), underscoring the need for workflows integrating minimal phenotypic confirmation.
Impact: Demonstrates real-world performance of a rapid blood culture panel in a low-resource setting with high neonatal burden and significant ESBL/carbapenemase prevalence, informing diagnostic stewardship and AMR surveillance.
Clinical Implications: Adopting BCID2 with targeted phenotypic confirmation can shorten time-to-identification and resistance reporting, enabling earlier optimization of empiric therapy and infection control in LMICs.
Key Findings
- Across 377 positive cultures with reference results, specificity exceeded 95% and sensitivity ranged from 50% (Acinetobacter calcoaceticus-baumannii complex, Proteus spp.) to 100% (Streptococcus pneumoniae, Salmonella spp.).
- CTX-M was detected in 111/175 (74.5%) Enterobacterales; 5/111 co-harbored NDM or VIM, with NDM-5 confirmed in 2/5 by sequencing.
- System usability was high (SUS 79.5), with a case-mix dominated by neonates (48.3%) and pediatric patients (39.8%).
Methodological Strengths
- Prospective validation with reference standards (MALDI-TOF or whole genome sequencing) and Wilson score confidence metrics.
- Parallel testing against standard phenotypic workflows and inclusion of a usability assessment.
Limitations
- Only 377/780 positives had reference results, potentially introducing selection bias.
- Sensitivity varied by organism, necessitating complementary phenotypic testing and workflow optimization.
Future Directions: Time-to-result and patient outcome studies, cost-effectiveness analyses, and expanded resistance target panels suited to local epidemiology.