Clinical Decision Support for Improved Neonatal Care: The Development of a Machine Learning Model for the Prediction of Late-onset Sepsis and Necrotizing Enterocolitis.

OBJECTIVE: To develop an artificial intelligence-based software system for predicting late-onset sepsis (LOS) and necrotizing enterocolitis (NEC) in infants admitted to the neonatal intensive care unit (NICU). STUDY DESIGN: Single-center, retrospective cohort study, conducted in the NICU of the Antwerp University Hospital. Continuous monitoring data of 865 preterm infants born at <32 weeks gestational age, admitted to the NICU in the first week of life, were used to train an XGBoost machine learning (ML) algorithm for LOS and NEC prediction in a cross-validated setup. Afterward, the model's performance was assessed on an independent test set of 148 patients (internal validation). RESULTS: The ML model delivered hourly risk predictions with an overall sensitivity of 69% (142/206) for all LOS/NEC episodes and 81% (67/83) for severe LOS/NEC episodes. The model showed a median time gain of ≤10 hours (IQR, 3.1-21.0 hours), compared with historical clinical diagnosis. On the complete retrospective dataset, the ML model made 721 069 predictions, of which 9805 (1.3%) depicted a LOS/NEC probability of ≥0.15, resulting in a total alarm rate of <1 patient alarm-day per week. The model reached a similar performance on the internal validation set. CONCLUSIONS: Artificial intelligence technology can assist clinicians in the early detection of LOS and NEC in the NICU, which potentially can result in clinical and socioeconomic benefits. Additional studies are required to quantify further the effect of combining artificial and human intelligence on patient outcomes in the NICU.

Machine Learning to Support Hemodynamic Intervention in the Neonatal Intensive Care Unit.

Hemodynamic support in neonatal intensive care is directed at maintaining cardiovascular wellbeing. At present, monitoring of vital signs plays an essential role in augmenting care in a reactive manner. By applying machine learning techniques, a model can be trained to learn patterns in time series data, allowing the detection of adverse outcomes before they become clinically apparent. In this review we provide an overview of the different machine learning techniques that have been used to develop models in hemodynamic care for newborn infants. We focus on their potential benefits, research pitfalls, and challenges related to their implementation in clinical care.

Compound Characterization and Metabolic Profile Elucidation after In Vitro Gastrointestinal and Hepatic Biotransformation of an Herniaria hirsuta Extract Using Unbiased Dynamic Metabolomic Data Analysis.

Herniaria hirsuta L. (Caryophyllaceae) is used for treatment of urinary stones and as a diuretic. Little is known about the active compounds and the mechanism of action. The phytochemical composition of H. hirsuta was comprehensively characterized using UHPLC-UV-HRMS (Ultrahigh-Performance Liquid Chromatography-Ultraviolet-High Resolution Mass Spectrometry) data. An in vitro gastrointestinal model was used to simulate biotransformation, which allowed the monitoring of the relative abundances of individual compounds over time. To analyze the longitudinal multiclass LC-MS data, XCMS, a platform that enables online metabolomics data processing and interpretation, and EDGE, a statistical method for time series data, were used to extract significant differential profiles from the raw data. An interactive Shiny app in R was used to rate the quality of the resulting features. These ratings were used to train a random forest model. The most abundant aglycone after gastrointestinal biotransformation was subjected to hepatic biotransformation using human S9 fractions. A diversity of compounds was detected, mainly saponins and flavonoids. Besides the known saponins, 15 new saponins were tentatively identified as glycosides of medicagenic acid, acetylated medicagenic acid and zanhic acid. It is suggested that metabolites of phytochemicals present in H. hirsuta, most likely saponins, are responsible for the pharmaceutical effects. It was observed that the relative abundance of saponin aglycones increased, indicating loss of sugar moieties during colonic biotransformation, with medicagenic acid as the most abundant aglycone. Hepatic biotransformation of this aglycone resulted in different metabolites formed by phase I and II reactions.

The use of plasma donor-derived, cell-free DNA to monitor acute rejection after kidney transplantation.

BACKGROUND: After transplantation, cell-free deoxyribonucleic acid (DNA) derived from the donor organ (ddcfDNA) can be detected in the recipient's circulation. We aimed to investigate the role of plasma ddcfDNA as biomarker for acute kidney rejection. METHODS: From 107 kidney transplant recipients, plasma samples were collected longitudinally after transplantation (Day 1 to 3 months) within a multicentre set-up. Cell-free DNA from the donor was quantified in plasma as a fraction of the total cell-free DNA by next generation sequencing using a targeted, multiplex polymerase chain reaction-based method for the analysis of single nucleotide polymorphisms. RESULTS: Increases of the ddcfDNA% above a threshold value of 0.88% were significantly associated with the occurrence of episodes of acute rejection (P = 0.017), acute tubular necrosis (P = 0.011) and acute pyelonephritis (P = 0.032). A receiver operating characteristic curve analysis revealed an equal area under the curve of the ddcfDNA% and serum creatinine of 0.64 for the diagnosis of acute rejection. CONCLUSIONS: Although increases in plasma ddcfDNA% are associated with graft injury, plasma ddcfDNA does not outperform the diagnostic capacity of the serum creatinine in the diagnosis of acute rejection.

Exposure of HepaRG Cells to Sodium Saccharin Underpins the Importance of Including Non-Hepatotoxic Compounds When Investigating Toxicological Modes of Action Using Metabolomics.

Metabolites represent the most downstream information of the cellular organisation. Hence, metabolomics experiments are extremely valuable to unravel the endogenous pathways involved in a toxicological mode of action. However, every external stimulus can introduce alterations in the cell homeostasis, thereby obscuring the involved endogenous pathways, biasing the interpretation of the results. Here we report on sodium saccharin, which is considered to be not hepatotoxic and therefore can serve as a reference compound to detect metabolic alterations that are not related to liver toxicity. Exposure of HepaRG cells to high levels of sodium saccharin (>10 mM) induced cell death, probably due to an increase in the osmotic pressure. Yet, a low number (n = 15) of significantly altered metabolites were also observed in the lipidome, including a slight decrease in phospholipids and an increase in triacylglycerols, upon daily exposure to 5 mM sodium saccharin for 72 h. The observation that a non-hepatotoxic compound can affect the metabolome underpins the importance of correct experimental design and data interpretation when investigating toxicological modes of action via metabolomics.

Untargeted liquid chromatography-mass spectrometry metabolomics to assess drug-induced cholestatic features in HepaRG® cells.

Cholestasis is a liver disease associated with retention of bile in the liver, which leads to local hepatic inflammation and severe liver damage. In order to investigate the mode of action of drug-induced cholestasis, in vitro models have shown to be able to recapitulate important elements of this disease. In this study, we applied untargeted metabolomics to investigate the metabolic perturbances in HepaRG® cells exposed for 24 h and 72 h to bosentan, a cholestatic reference toxicant. Intracellular profiles were extracted and analysed with liquid chromatography and accurate-mass spectrometry. Metabolites of interest were selected using partial least-squares discriminant analysis and random forest classifier models. The observed metabolic patterns associated with cholestasis in vitro were complex. Acute (24 h) exposure revealed metabolites related to apoptosis, such as ceramide and triglyceride accumulation, in combination with phosphatidylethanolamine, choline and carnitine depletion. Metabolomic alterations during exposure to lower dosages and a prolonged exposure (72 h) included carnitine upregulation and changes in the polyamine metabolism. These metabolites were linked to changes in phospholipid metabolism, mitochondrial pathways and energy homeostasis. The metabolic changes confirmed the mitotoxic effects of bosentan and revealed the potential involvement of phospholipid metabolism as part of the mode of action of drug-induced cholestasis.

Using Expert Driven Machine Learning to Enhance Dynamic Metabolomics Data Analysis.

Data analysis for metabolomics is undergoing rapid progress thanks to the proliferation of novel tools and the standardization of existing workflows. As untargeted metabolomics datasets and experiments continue to increase in size and complexity, standardized workflows are often not sufficiently sophisticated. In addition, the ground truth for untargeted metabolomics experiments is intrinsically unknown and the performance of tools is difficult to evaluate. Here, the problem of dynamic multi-class metabolomics experiments was investigated using a simulated dataset with a known ground truth. This simulated dataset was used to evaluate the performance of tinderesting, a new and intuitive tool based on gathering expert knowledge to be used in machine learning. The results were compared to EDGE, a statistical method for time series data. This paper presents three novel outcomes. The first is a way to simulate dynamic metabolomics data with a known ground truth based on ordinary differential equations. This method is made available through the MetaboLouise R package. Second, the EDGE tool, originally developed for genomics data analysis, is highly performant in analyzing dynamic case vs. control metabolomics data. Third, the tinderesting method is introduced to analyse more complex dynamic metabolomics experiments. This tool consists of a Shiny app for collecting expert knowledge, which in turn is used to train a machine learning model to emulate the decision process of the expert. This approach does not replace traditional data analysis workflows for metabolomics, but can provide additional information, improved performance or easier interpretation of results. The advantage is that the tool is agnostic to the complexity of the experiment, and thus is easier to use in advanced setups. All code for the presented analysis, MetaboLouise and tinderesting are freely available.

Revelation of the metabolic pathway of hederacoside C using an innovative data analysis strategy for dynamic multiclass biotransformation experiments.

Although some herbal remedies have been used for decades, little is known about the active compounds and the mechanism of action. Many natural products, such as glycosides, can be considered as prodrugs, which become active after biotransformation. To optimize the workflow of in vitro biotransformation followed by automated data analysis, hederacoside C was used as a model compound for saponins. Hederacoside C was subjected to gastrointestinal enzymes and fecal microflora. Samples were analyzed with UHPLC-PDA-HRMS before, during and after in vitro biotransformation, which allowed the monitoring of the relative abundances of the compound and its metabolites. The data-analysis workflow was optimized to render as much information as possible from the longitudinal LCMS data. XCMS was used to convert the raw data into features via peak-picking, followed by grouping, and EDGE was used for the extraction of significant differential profiles. To evaluate if the workflow was suitable for dynamic multiclass metabolomics data, an interactive Shiny web app was developed in R to rate the quality of the resulting features. These ratings were used to train a random forest model for predicting experts response. A performance analysis revealed that the random forest model was capable of correctly predicting the reviewers response in most cases (AUC 0.926 with 10 fold cross validation). The automated data analysis workflow was used for unbiased screening for metabolites and revealed the biotransformation of hederacoside C. As expected, a decrease in relative abundance of hederacoside C was observed over time. Additionally, the relative abundance of metabolites increased, illustrating the biotransformation of hederacoside C, especially in the colon phase, where microbial fermentation takes place. Stepwise progressive elimination of sugar moieties was the major metabolic pathway.

Plasma donor-derived cell-free DNA kinetics after kidney transplantation using a single tube multiplex PCR assay.

BACKGROUND: After transplantation, cell-free DNA derived from the donor organ (ddcfDNA) can be detected in the recipient's circulation. We aimed to quantify ddcfDNA levels in plasma of kidney transplant recipients thereby investigating the kinetics of this biomarker after transplantation and determining biological variables that influence ddcfDNA kinetics in stable and non-stable patients. MATERIALS AND METHODS: From 107 kidney transplant recipients, plasma samples were collected longitudinally after transplantation (day 1-3 months) within a multicenter set-up. Cell-free DNA from the donor was quantified in plasma as a fraction of the total cell-free DNA by next generation sequencing using a targeted, multiplex PCR-based method for the analysis of single nucleotide polymorphisms. A subgroup of stable renal transplant recipients was identified to determine a ddcfDNA threshold value. RESULTS: In stable transplant recipients, plasma ddcfDNA% decreased to a mean (SD) ddcfDNA% of 0.46% (± 0.21%) which was reached 9.85 (± 5.6) days after transplantation. A ddcfDNA threshold value of 0.88% (mean + 2SD) was determined in kidney transplant recipients. Recipients that did not reach this threshold ddcfDNA value within 10 days after transplantation showed a higher ddcfDNA% on the first day after transplantation and demonstrated a higher individual baseline ddcfDNA%. CONCLUSION: In conclusion, plasma ddcfDNA fractions decreased exponentially within 10 days after transplantation to a ddcfDNA threshold value of 0.88% or less. To investigate the role of ddcfDNA for rejection monitoring of the graft, future research is needed to determine causes of ddcfDNA% increases above this threshold value.

speaq 2.0: A complete workflow for high-throughput 1D NMR spectra processing and quantification.

Nuclear Magnetic Resonance (NMR) spectroscopy is, together with liquid chromatography-mass spectrometry (LC-MS), the most established platform to perform metabolomics. In contrast to LC-MS however, NMR data is predominantly being processed with commercial software. Meanwhile its data processing remains tedious and dependent on user interventions. As a follow-up to speaq, a previously released workflow for NMR spectral alignment and quantitation, we present speaq 2.0. This completely revised framework to automatically analyze 1D NMR spectra uses wavelets to efficiently summarize the raw spectra with minimal information loss or user interaction. The tool offers a fast and easy workflow that starts with the common approach of peak-picking, followed by grouping, thus avoiding the binning step. This yields a matrix consisting of features, samples and peak values that can be conveniently processed either by using included multivariate statistical functions or by using many other recently developed methods for NMR data analysis. speaq 2.0 facilitates robust and high-throughput metabolomics based on 1D NMR but is also compatible with other NMR frameworks or complementary LC-MS workflows. The methods are benchmarked using a simulated dataset and two publicly available datasets. speaq 2.0 is distributed through the existing speaq R package to provide a complete solution for NMR data processing. The package and the code for the presented case studies are freely available on CRAN (https://cran.r-project.org/package=speaq) and GitHub (https://github.com/beirnaert/speaq).

On the feasibility of mining CD8+ T cell receptor patterns underlying immunogenic peptide recognition.

Current T cell epitope prediction tools are a valuable resource in designing targeted immunogenicity experiments. They typically focus on, and are able to, accurately predict peptide binding and presentation by major histocompatibility complex (MHC) molecules on the surface of antigen-presenting cells. However, recognition of the peptide-MHC complex by a T cell receptor (TCR) is often not included in these tools. We developed a classification approach based on random forest classifiers to predict recognition of a peptide by a T cell receptor and discover patterns that contribute to recognition. We considered two approaches to solve this problem: (1) distinguishing between two sets of TCRs that each bind to a known peptide and (2) retrieving TCRs that bind to a given peptide from a large pool of TCRs. Evaluation of the models on two HIV-1, B*08-restricted epitopes reveals good performance and hints towards structural CDR3 features that can determine peptide immunogenicity. These results are of particular importance as they show that prediction of T cell epitope and T cell epitope recognition based on sequence data is a feasible approach. In addition, the validity of our models not only serves as a proof of concept for the prediction of immunogenic T cell epitopes but also paves the way for more general and high-performing models.

The future of metabolomics in ELIXIR.

Metabolomics, the youngest of the major omics technologies, is supported by an active community of researchers and infrastructure developers across Europe. To coordinate and focus efforts around infrastructure building for metabolomics within Europe, a workshop on the "Future of metabolomics in ELIXIR" was organised at Frankfurt Airport in Germany. This one-day strategic workshop involved representatives of ELIXIR Nodes, members of the PhenoMeNal consortium developing an e-infrastructure that supports workflow-based metabolomics analysis pipelines, and experts from the international metabolomics community. The workshop established metabolite identification as the critical area, where a maximal impact of computational metabolomics and data management on other fields could be achieved. In particular, the existing four ELIXIR Use Cases, where the metabolomics community - both industry and academia - would benefit most, and which could be exhaustively mapped onto the current five ELIXIR Platforms were discussed. This opinion article is a call for support for a new ELIXIR metabolomics Use Case, which aligns with and complements the existing and planned ELIXIR Platforms and Use Cases.

Tailored liquid chromatography-mass spectrometry analysis improves the coverage of the intracellular metabolome of HepaRG cells.

Metabolomics protocols are often combined with Liquid Chromatography-Mass Spectrometry (LC-MS) using mostly reversed phase chromatography coupled to accurate mass spectrometry, e.g. quadrupole time-of-flight (QTOF) mass spectrometers to measure as many metabolites as possible. In this study, we optimised the LC-MS separation of cell extracts after fractionation in polar and non-polar fractions. Both phases were analysed separately in a tailored approach in four different runs (two for the non-polar and two for the polar-fraction), each of them specifically adapted to improve the separation of the metabolites present in the extract. This approach improves the coverage of a broad range of the metabolome of the HepaRG cells and the separation of intra-class metabolites. The non-polar fraction was analysed using a C18-column with end-capping, mobile phase compositions were specifically adapted for each ionisation mode using different co-solvents and buffers. The polar extracts were analysed with a mixed mode Hydrophilic Interaction Liquid Chromatography (HILIC) system. Acidic metabolites from glycolysis and the Krebs cycle, together with phosphorylated compounds, were best detected with a method using ion pairing (IP) with tributylamine and separation on a phenyl-hexyl column. Accurate mass detection was performed with the QTOF in MS-mode only using an extended dynamic range to improve the quality of the dataset. Parameters with the greatest impact on the detection were the balance between mass accuracy and linear range, the fragmentor voltage, the capillary voltage, the nozzle voltage, and the nebuliser pressure. By using a tailored approach for the intracellular HepaRG metabolome, consisting of three different LC techniques, over 2200 metabolites can be measured with a high precision and acceptable linear range. The developed method is suited for qualitative untargeted LC-MS metabolomics studies.

Asymmetric dyes align inside carbon nanotubes to yield a large nonlinear optical response.

Asymmetric dye molecules have unusual optical and electronic properties. For instance, they show a strong second-order nonlinear optical (NLO) response that has attracted great interest for potential applications in electro-optic modulators for optical telecommunications and in wavelength conversion of lasers. However, the strong Coulombic interaction between the large dipole moments of these molecules favours a pairwise antiparallel alignment that cancels out the NLO response when incorporated into bulk materials. Here, we show that by including an elongated dipolar dye (p,p'-dimethylaminonitrostilbene, DANS, a prototypical asymmetric dye with a strong NLO response) inside single-walled carbon nanotubes (SWCNTs), an ideal head-to-tail alignment in which all electric dipoles point in the same sense is naturally created. We have applied this concept to synthesize solution-processible DANS-filled SWCNTs that show an extremely large total dipole moment and static hyperpolarizability (ß0 = 9,800 × 10(-30) e.s.u.), resulting from the coherent alignment of arrays of ∼70 DANS molecules.