Browse Articles
Discover research articles across all indexed journals
Achieving flexible fairness metrics in federated medical imaging
Identifying early predictive and diagnostic biomarkers and exploring metabolic pathways for sepsis after trauma based on an untargeted metabolomics approach
Abstract Systemic inflammatory response syndrome (SIRS) and organ dysfunction make it challenging to predict which major trauma patients are at risk of developing sepsis. Additionally, the unclear pathogenesis of sepsis after trauma contributes to its high morbidity and mortality. Identifying early predictive and diagnostic biomarkers, as well as exploring related metabolic pathways, is crucial for improving early prevention, diagnosis, and treatment. This study prospectively analyzed plasma samples from patients with severe trauma collected between March 2022 and November 2023. Trauma patients were divided into two groups based on whether they developed sepsis within two weeks: the TDDS group (trauma patients who did not develop sepsis) and the TDS group (trauma patients who did develop sepsis). Plasma samples from the TDS group were collected at the time of sepsis diagnosis (Sepsis group). Metabolite concentrations were measured using ultrahigh-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) through untargeted metabolomics. From the differential metabolites between the TDS and TDDS groups, we identified five significant metabolites (all area under the curve (AUC) ≥ 0.94) as early predictive biomarkers for sepsis after trauma: (1) docosatrienoic acid, (2) 7-alpha-carboxy-17-alpha-carboxyethylandrostan lactone phenyl ester, (3) sphingomyelin (SM) 8:1;2O/26:1, (4) N1-[1-(3-isopropenylphenyl)-1-methylethyl]-3-oxobutanamide, and (5) SM 34:2;2O. Furthermore, five significant metabolites (all AUC ≥ 0.85) were identified as early diagnostic biomarkers from the comparison between the TDS and TDDS groups: (1) lysophosphatidylcholine (LPC) O-22:1, (2) LPC O-22:0, (3) uric acid, (4) LPC O-24:2, and (5) LPC 22:0-SN1. 26 metabolites shared between two comparisons (TDS vs. TDDS and sepsis vs. TDS) were identified. Of which, 19 metabolites belong to lipid metabolism. The top three metabolic pathways related to sepsis after trauma under the impact of severe trauma were: (1) glycerophospholipid metabolism, (2) porphyrin metabolism, and (3) sphingolipid metabolism. The top three metabolic pathways related to sepsis after trauma under the impact of infection were: (1) caffeine metabolism, (2) biosynthesis of unsaturated fatty acids, and (3) steroid hormone biosynthesis. Our study identified early predictive and diagnostic biomarkers and explored metabolic pathways related to sepsis after trauma. These findings provide a foundation for future research on the onset and development of sepsis, facilitating its early prevention, diagnosis, and treatment based on specific metabolites and metabolic pathways.
A naturally selected αβ T cell receptor binds HLA-DQ2 molecules without co-contacting the presented peptide
Abstract αβ T cell receptors (TCR) co-recognise peptide (p) antigens that are presented by major histocompatibility complex (MHC) molecules. While marked variations in TCR-p-MHC docking topologies have been observed from structural studies, the co-recognition paradigm has held fast. Using HLA-DQ2.5-peptide tetramers, here we identify a TRAV12-1+-TRBV5-1+ G9 TCR from human peripheral blood that binds HLA-DQ2.5 in a peptide-agnostic manner. The crystal structures of TCR-HLA-DQ2.5-peptide complexes show that the G9 TCR binds HLA-DQ2.5 in a reversed docking topology without contacting the peptide, with the TCR contacting the β1 region of HLA-DQ2.5 and distal from the peptide antigen binding cleft. High-throughput screening of HLA class I and II molecules finds the G9 TCR to be pan-HLA-DQ2 reactive, with leucine-55 of HLA-DQ2.5 being a key determinant underpinning G9 TCR specificity excluding other HLA-II allomorphs. Consistent with the functional assays, the interactions of the G9 TCR and HLA-DQ2.5 precludes CD4 binding, thereby impeding T cell activation. Collectively, we describe a naturally selected αβTCR from human peripheral blood that deviates from the TCR-p-MHC co-recognition paradigm.
The synergistic activity of fisetin on quercetin improves testicular recover in ischemia-reperfusion injury in rats
Data splitting to avoid information leakage with DataSAIL
Abstract Information leakage is an increasingly important topic in machine learning research for biomedical applications. When information leakage happens during a model’s training, it risks memorizing the training data instead of learning generalizable properties. This can lead to inflated performance metrics that do not reflect the actual performance at inference time. We present DataSAIL, a versatile Python package to facilitate leakage-reduced data splitting to enable realistic evaluation of machine learning models for biological data that are intended to be applied in out-of-distribution scenarios. DataSAIL is based on formulating the problem to find leakage-reduced data splits as a combinatorial optimization problem. We prove that this problem is NP-hard and provide a scalable heuristic based on clustering and integer linear programming. Finally, we empirically demonstrate DataSAIL’s impact on evaluating biomedical machine learning models.