Early Detection of Septic Shock Onset Using Interpretable Machine Learners.

BACKGROUND: Developing a decision support system based on advances in machine learning is one area for strategic innovation in healthcare. Predicting a patient's progression to septic shock is an active field of translational research. The goal of this study was to develop a working model of a clinical decision support system for predicting septic shock in an acute care setting for up to 6 h from the time of admission in an integrated healthcare setting. METHOD: Clinical data from Electronic Health Record (EHR), at encounter level, were used to build a predictive model for progression from sepsis to septic shock up to 6 h from the time of admission; that is, T = 1, 3, and 6 h from admission. Eight different machine learning algorithms (Random Forest, XGBoost, C5.0, Decision Trees, Boosted Logistic Regression, Support Vector Machine, Logistic Regression, Regularized Logistic, and Bayes Generalized Linear Model) were used for model development. Two adaptive sampling strategies were used to address the class imbalance. Data from two sources (clinical and billing codes) were used to define the case definition (septic shock) using the Centers for Medicare & Medicaid Services (CMS) Sepsis criteria. The model assessment was performed using Area under Receiving Operator Characteristics (AUROC), sensitivity, and specificity. Model predictions for each feature window (1, 3 and 6 h from admission) were consolidated. RESULTS: Retrospective data from April 2005 to September 2018 were extracted from the EHR, Insurance Claims, Billing, and Laboratory Systems to create a dataset for septic shock detection. The clinical criteria and billing information were used to label patients into two classes-septic shock patients and sepsis patients at three different time points from admission, creating two different case-control cohorts. Data from 45,425 unique in-patient visits were used to build 96 prediction models comparing clinical-based definition versus billing-based information as the gold standard. Of the 24 consolidated models (based on eight machine learning algorithms and three feature windows), four models reached an AUROC greater than 0.9. Overall, all the consolidated models reached an AUROC of at least 0.8820 or higher. Based on the AUROC of 0.9483, the best model was based on Random Forest, with a sensitivity of 83.9% and specificity of 88.1%. The sepsis detection window at 6 h outperformed the 1 and 3-h windows. The sepsis definition based on clinical variables had improved performance when compared to the sepsis definition based on only billing information. CONCLUSION: This study corroborated that machine learning models can be developed to predict septic shock using clinical and administrative data. However, the use of clinical information to define septic shock outperformed models developed based on only administrative data. Intelligent decision support tools can be developed and integrated into the EHR and improve clinical outcomes and facilitate the optimization of resources in real-time.

Artificial Intelligence Transforms the Future of Health Care.

Life sciences researchers using artificial intelligence (AI) are under pressure to innovate faster than ever. Large, multilevel, and integrated data sets offer the promise of unlocking novel insights and accelerating breakthroughs. Although more data are available than ever, only a fraction is being curated, integrated, understood, and analyzed. AI focuses on how computers learn from data and mimic human thought processes. AI increases learning capacity and provides decision support system at scales that are transforming the future of health care. This article is a review of applications for machine learning in health care with a focus on clinical, translational, and public health applications with an overview of the important role of privacy, data sharing, and genetic information.

NLRX1 Modulates Immunometabolic Mechanisms Controlling the Host-Gut Microbiota Interactions during Inflammatory Bowel Disease.

Interactions among the gut microbiome, dysregulated immune responses, and genetic factors contribute to the pathogenesis of inflammatory bowel disease (IBD). Nlrx1-/- mice have exacerbated disease severity, colonic lesions, and increased inflammatory markers. Global transcriptomic analyses demonstrate enhanced mucosal antimicrobial defense response, chemokine and cytokine expression, and epithelial cell metabolism in colitic Nlrx1-/- mice compared to wild-type (WT) mice. Cell-specificity studies using cre-lox mice demonstrate that the loss of NLRX1 in intestinal epithelial cells (IEC) recapitulate the increased sensitivity to DSS colitis observed in whole body Nlrx1-/- mice. Further, organoid cultures of Nlrx1-/- and WT epithelial cells confirm the altered patterns of proliferation, amino acid metabolism, and tight junction expression. These differences in IEC behavior can impact the composition of the microbiome. Microbiome analyses demonstrate that colitogenic bacterial taxa such as Veillonella and Clostridiales are increased in abundance in Nlrx1-/- mice and in WT mice co-housed with Nlrx1-/- mice. The transfer of an Nlrx1-/--associated gut microbiome through co-housing worsens disease in WT mice confirming the contributions of the microbiome to the Nlrx1-/- phenotype. To validate NLRX1 effects on IEC metabolism mediate gut-microbiome interactions, restoration of WT glutamine metabolic profiles through either exogenous glutamine supplementation or administration of 6-diazo-5-oxo-l-norleucine abrogates differences in inflammation, microbiome, and overall disease severity in Nlrx1-/- mice. The influence NLRX1 deficiency on SIRT1-mediated effects is identified to be an upstream controller of the Nlrx1-/- phenotype in intestinal epithelial cell function and metabolism. The altered IEC function and metabolisms leads to changes in barrier permeability and microbiome interactions, in turn, promoting greater translocation and inflammation and resulting in an increased disease severity. In conclusion, NLRX1 is an immunoregulatory molecule and a candidate modulator of the interplay between mucosal inflammation, metabolism, and the gut microbiome during IBD.

Modeling-Enabled Characterization of Novel NLRX1 Ligands.

Nucleotide-binding domain and leucine-rich repeat containing (NLR) family are intracellular sentinels of cytosolic homeostasis that orchestrate immune and inflammatory responses in infectious and immune-mediated diseases. NLRX1 is a mitochondrial-associated NOD-like receptor involved in the modulation of immune and metabolic responses. This study utilizes molecular docking approaches to investigate the structure of NLRX1 and experimentally assesses binding to naturally occurring compounds from several natural product and lipid databases. Screening of compound libraries predicts targeting of NLRX1 by conjugated trienes, polyketides, prenol lipids, sterol lipids, and coenzyme A-containing fatty acids for activating the NLRX1 pathway. The ligands of NLRX1 were identified by docking punicic acid (PUA), eleostearic acid (ESA), and docosahexaenoic acid (DHA) to the C-terminal fragment of the human NLRX1 (cNLRX1). Their binding and that of positive control RNA to cNLRX1 were experimentally determined by surface plasmon resonance (SPR) spectroscopy. In addition, the ligand binding sites of cNLRX1 were predicted in silico and validated experimentally. Target mutagenesis studies demonstrate that mutation of 4 critical residues ASP677, PHE680, PHE681, and GLU684 to alanine resulted in diminished affinity of PUA, ESA, and DHA to NLRX1. Consistent with the regulatory actions of NLRX1 on the NF-κB pathway, treatment of bone marrow derived macrophages (BMDM)s with PUA and DHA suppressed NF-κB activity in a NLRX1 dependent mechanism. In addition, a series of pre-clinical efficacy studies were performed using a mouse model of dextran sodium sulfate (DSS)-induced colitis. Our findings showed that the regulatory function of PUA on colitis is NLRX1 dependent. Thus, we identified novel small molecules that bind to NLRX1 and exert anti-inflammatory actions.

An automated framework for hypotheses generation using literature.

BACKGROUND: In bio-medicine, exploratory studies and hypothesis generation often begin with researching existing literature to identify a set of factors and their association with diseases, phenotypes, or biological processes. Many scientists are overwhelmed by the sheer volume of literature on a disease when they plan to generate a new hypothesis or study a biological phenomenon. The situation is even worse for junior investigators who often find it difficult to formulate new hypotheses or, more importantly, corroborate if their hypothesis is consistent with existing literature. It is a daunting task to be abreast with so much being published and also remember all combinations of direct and indirect associations. Fortunately there is a growing trend of using literature mining and knowledge discovery tools in biomedical research. However, there is still a large gap between the huge amount of effort and resources invested in disease research and the little effort in harvesting the published knowledge. The proposed hypothesis generation framework (HGF) finds "crisp semantic associations" among entities of interest - that is a step towards bridging such gaps. METHODOLOGY: The proposed HGF shares similar end goals like the SWAN but are more holistic in nature and was designed and implemented using scalable and efficient computational models of disease-disease interaction. The integration of mapping ontologies with latent semantic analysis is critical in capturing domain specific direct and indirect "crisp" associations, and making assertions about entities (such as disease X is associated with a set of factors Z). RESULTS: Pilot studies were performed using two diseases. A comparative analysis of the computed "associations" and "assertions" with curated expert knowledge was performed to validate the results. It was observed that the HGF is able to capture "crisp" direct and indirect associations, and provide knowledge discovery on demand. CONCLUSIONS: The proposed framework is fast, efficient, and robust in generating new hypotheses to identify factors associated with a disease. A full integrated Web service application is being developed for wide dissemination of the HGF. A large-scale study by the domain experts and associated researchers is underway to validate the associations and assertions computed by the HGF.