Diabetes Spotlight: Kevin Williams, PhD-Mapping the Neural Networks of Metabolic Systems.

A Hybrid Input-Output Approach to Model Metabolic Systems: An Application to Intracellular Thiamine Kinetics

The Cardiopulmononary Exercise Test (CPET) is a diagnostic test that evaluates the functional capacity of na individual through the integrated response of the cardiovascular, respiratory and metabolic systems. VO₂max is the parameter that acess functional capacity, although it’s difficult to achieve given the effort that implies. In recent years, an increase in computing capabilities combined with available storage of large amounts of information has led to a heightened interest in machine learning (ML). We aimed in this study to enable CPET with ML models that allow predicting oxygen consumption in healthy individuals. The study methodology is based on the cleaning and exploratory analysis of a public database with about 992 CPET performed on healthy individuals and athletes. To predict the each value of VO₂ (~569,000 instances), five ML algorithms were used (Random Forests, kNN, Neural Networks, Linear Regression and SVM) with heart rate, respiratory rate, time from the beginning of the exame and treadmill speed, using a 20-fold cross-validation. The best result came from the Random Forest model, with a R² of 0.88 and a RMSE of 334.34 ml.min^-1. Futhermore, using the same methodology but different features, we tried to predict the the VO₂max with the 724 adult participants with a maximal test (RER≥1.05) but weaker results were obtained (best model was the Linear Regression, with a R² of 0.50 and a RMSE of 498.06 ml.min^-1). Still, this model showed a better correlation with the real VO₂max than the Wasserman equation (R=0.71 vs R=0.59). It’s possible to predict with accuracy breath-by-breath VO₂, based in easy-to-obtain physiological and mechanical measurements.

Mechanistic and therapeutic relationships of traumatic brain injury and γ-amino-butyric acid (GABA)

The emergence and evolution of worldviews is a complex phenomenon that requires strong and rigorous scientific attention in our hyperconnected world. On the one hand, cognitive theories have proposed reasonable frameworks but have not reached general modeling frameworks where predictions can be tested. On the other hand, machine-learning-based applications perform extremely well at predicting outcomes of worldviews, but they rely on a set of optimized weights in a neural network that does not comply to a well-founded cognitive framework. In this article, we propose a formal approach used to investigate the establishment of and change in worldviews by recalling that the realm of ideas, where opinions, perspectives and worldviews are shaped, resemble, in many ways, a metabolic system. We propose a general modelization of worldviews based on reaction networks, and a specific starting model based on species representing belief attitudes and species representing belief change triggers. These two kinds of species combine and modify their structures through the reactions. We show that chemical organization theory combined with dynamical simulations can illustrate various interesting features of how worldviews emerge, are maintained and change. In particular, worldviews correspond to chemical organizations, meaning closed and self-producing structures, which are generally maintained by feedback loops occurring within the beliefs and triggers in the organization. We also show how, by inducing the external input of belief change triggers, it is possible to change from one worldview to another, in an irreversible way. We illustrate our approach with a simple example reflecting the formation of an opinion and a belief attitude about a theme, and, next, show a more complex scenario containing opinions and belief attitudes about two possible themes.

Gray matters—the really big picture

Biochemical neuron: hardware implementation of functional devices by mimicking the natural intelligence such as metabolic control systems

Diabetes is a disease characterized by a high level of sugar in the blood. The disease occurs because of a disruption in the metabolic system when insulin is not produced effectively and functions properly. High blood sugar levels, for an extended period of time, can harm a few organ systems, including the heart and kidneys. Moreover, it may cause blindness or death if it is not carefully monitored. Because diabetes symptoms are rarely seen, one of the factors that may cause diabetes is self-awareness. Thus, with Artificial Intelligence, this problem can be solved. Artificial intelligence studies how machines can function like humans. This study implemented a Convolutional Neural Network algorithm with (1) input layer, (2) feature learning layer, (3) classification layer, and (4) output layer as the architecture for AI. The accuracy of the developed AI model was measured from its precision, recall, and f1-score. The results show that the model obtained 90% precision, recall, and f1-score for real-world cases found in two hospitals located in Solo and Yogyakarta, Indonesia. According to the results of the tests, 9 out of 10 patients were correctly predicted as having a high risk of diabetes based on their eye images.

In this monograph, we have developed a dynamical systems and control theory framework for network information systems. These systems are composed of networked interconnected large-scale systems and include air traffic control systems, power and energy grid systems, manufacturing and processing systems, aerospace and transportation systems, communication and information networks, integrative biological systems, biological neural networks, biomolecular and biochemical systems, nervous systems, immune systems, environmental and ecological systems, molecular and chemical reaction systems, economic and financial systems, cellular systems, metabolic systems, and ecosystems, to name but a few examples.

Total brachial plexus injuries (TBPI) caused by trauma often inflict unbearable pain on patients. However, there is still a lack of effective treatment strategies for these conditions. Optogenetics, as an emerging interdisciplinary field, can bridge biology and electronic science, becoming an important research tool for studying biological neural networks and metabolic systems. In this study, we fabricate a flexible optogenetic probe through micro-LED technology which is able to directly stimulate photosensitive proteins via the high brightness of micro-LEDs, effectively modulating neuronal activity. The flexible probe has excellent stability and optical properties. Under 5 mm bending, there is almost no obvious change in optical properties. Furthermore, with a duty cycle of 10%, the temperature increase is only 0.6 °C. This low temperature change can effectively prevent the brain damage caused by its thermal effect. We demonstrate that this approach can significantly ameliorate pain associated with TBPI.

The upper limb rehabilitation robot is in direct contact with the patient, so the control of the robot needs to be compliant. In this paper, an impedance control method of three degree of freedom upper limb rehabilitation robot is proposed. In the impedance control model, the force between the patients and machine is simplified into a spring damping system. The neural network PID is used to control the robot to achieve the desired trajectory. The learning function of neural network can adjust parameters of PID online which enables the impedance control method to control the metabolic system. Finally, the test results show that the impedance control method has certain superiority.

High-throughput biological data, thanks to technological advances, have become cheaper to collect, leading to the availability of vast amounts of omic data of different types. In parallel, the in silico reconstruction and modeling of metabolic systems is now acknowledged as a key tool to complement experimental data on a large scale. The integration of these model- and data-driven information is therefore emerging as a new challenge in systems biology, with no clear guidance on how to better take advantage of the inherent multisource and multiomic nature of these data types while preserving mechanistic interpretation. Here, we investigate different regularization techniques for high-dimensional data derived from the integration of gene expression profiles with metabolic flux data, extracted from strain-specific metabolic models, to improve cellular growth rate predictions. To this end, we propose ad-hoc extensions of previous regularization frameworks including group, view-specific and principal component regularization and experimentally compare them using data from 1143 Saccharomyces cerevisiae strains. We observe a divergence between methods in terms of regression accuracy and integration effectiveness based on the type of regularization employed. In multiomic regression tasks, when learning from experimental and model-generated omic data, our results demonstrate the competitiveness and ease of interpretation of multimodal regularized linear models compared to data-hungry methods based on neural networks. All data, models and code produced in this work are available on GitHub at https://github.com/Angione-Lab/HybridGroupIPFLasso_pc2Lasso. Supplementary data are available at Bioinformatics online.

The “structural unit” of the vertebrate brain has been identified as a multi-cellular unit, formed from a single neuron and at least one each of its associated macroglial chaperone and vascular endothelial cells, which exhibits most of the fundamental properties of a brain. These properties include its genetic codes, cellular associations, mechanisms of neuronal information encoding, and its “operating system”, a homeostatic energy supply mechanism that enables neurons to continuously communicate with other neurons regardless of the bioenergetic demands made on the neural network in any part of the brain at any time. A structural unit is defined by three cell types required for the unique tri-cellular metabolism of N-acetylaspartate (NAA) and N-acetylaspartylglutamate (NAAG) including neurons, oligodendrocytes and astrocytes, and by their physiological roles operating in a four-cell domain which also includes vascular endothelial cells. A “functional unit” of the brain is a two-neuron entity, defined by the minimum number of neurons required for rapid intercellular communication. Thus, each functional unit is formed by the process of synaptogenesis from two single-neuron structural units and represents the smallest unit that exhibits all of the basic signaling properties present in a complex brain. These properties include all mechanisms of neuronal connectivity, information storage, and signaling. Since a structural unit is defined by NAA and NAAG intercellular metabolism and physiology, and two structural units form a functional unit, the NAA-NAAG system is intimately associated with all normal brain activities as well as all brain disorders. In this review, the hierarchal structural and functional units of the brain are described. In addition, a bioenergetic rationale for using the NAA-NAAG metabolic system as a biomarker for neuronal abundance and/or viability is presented, and examples of some human diseases that can be traced to interference with one or more components of structural or functional units are provided.

Metabolic P systems, also called MP systems, are discrete dynamical systems which proved to be effective for modeling biological systems. Their dynamics is generated by means of a metabolic algorithm based on flux regulation A significant problem related to the generation of MP models from experimental data concerns the synthesis of these functions. In this paper we introduce a new approach to the synthesis of MP fluxes relying on neural networks as universal function approximators, and on evolutionary algorithms as learning techniques. This methodology is successfully tested in the case study of mitotic oscillator in early amphibian embryos.

Στην παρούσα διατριβή σχεδιάζονται, αναπτύσσονται και αξιολογούνται ευφυή συστήματα υποστήριξης εξατομικευμένων ιατρικών αποφάσεων που στοχεύουν στη βελτιστοποίηση της θεραπείας των ατόμων με Σακχαρώδη Διαβήτη (ΣΔ). Συγκεκριμένα, οι μέθοδοι που αναπτύσσονται χρησιμοποιούνται για την ανάλυση και την επεξεργασία δεδομένων Ηλεκτρονικού Ιατρικού Φακέλου, Εργαστηριακών Μετρήσεων καθώς και συνεχών καταγραφών γλυκόζης και ινσουλίνης, με σκοπό i) τη σχεδίαση και ανάπτυξη Συμβουλευτικού Συστήματος Έγχυσης Ινσουλίνης (ΣΣΕΙ), το οποίο εκτιμά σε πραγματικό χρόνο τον απαιτούμενο ρυθμό έγχυσης ινσουλίνης σε άτομα με ΣΔ Τύπου Ι, που χρησιμοποιούν Διάταξη Συνεχούς Μέτρησης Γλυκόζης (ΔΣΜΓ) και αντλία έγχυσης ινσουλίνης («Τεχνητό Πάγκρεας»), ώστε τα επίπεδα γλυκόζης αίματος, να διατηρούνται εντός φυσιολογικών ορίων και ii) την ανάπτυξη μοντέλων αποτίμησης της πιθανότητας εμφάνισης μακροπρόθεσμων επιπλοκών του ΣΔ Τύπου Ι και Τύπου ΙΙ, εστιάζοντας στη διαβητική αμφιβληστροειδοπάθεια. Στο πρώτο μέρος της διατριβής εφαρμόζονται προηγμένες μέθοδοι μοντελοποίησης, που βασίζονται στη συνδυασμένη χρήση Διαμερισματικών Μοντέλων (ΔΜ) και Νευρωνικών Δικτύων (ΝΔ) για την προσομοίωση του μεταβολικού συστήματος γλυκόζης-ινσουλίνης σε άτομα με ΣΔ Τύπου Ι. Το τελικό μοντέλο ενσωματώνεται σε έναν ελεγκτή που βασίζεται σε μοντέλο πρόβλεψης (Model Predictive Control-MPC), για τον μετέπειτα υπολογισμό των συνιστώμενων ρυθμών έγχυσης ινσουλίνης. Για τον λεπτομερή έλεγχο ορθής λειτουργίας του ΣΣΕΙ, πραγματοποιήθηκε σειρά υπολογιστικών πειραμάτων. Επιπλέον, διεξήχθη κλινική δοκιμή σε νοσοκομείο υπό ελεγχόμενες συνθήκες, τα αποτελέσματα της οποίας ανέδειξαν αδυναμίες του ΣΣΕΙ και οδήγησαν στη βελτίωσή του. Συγκεκριμένα, αναπτύχθηκε προσαρμοστικός αλγόριθμος αυτόματης και σε πραγματικό χρόνο, ενημέρωσης των παραμέτρων του ελεγκτή χρησιμοποιώντας τεχνικές ασαφούς λογικής. Το βελτιωμένο ΣΣΕΙ εξετάστηκε ως προς την ικανότητά του να διαχειρίζεται διαταραχές γευμάτων, καταστάσεις νηστείας, καθυστερήσεις, ανακρίβειες στις μετρήσεις γλυκόζης, διαφορές στο μεταβολισμό γλυκόζης που υφίστανται μεταξύ ατόμων με ΣΔ Τύπου Ι (inter-patient variability), καθώς και λανθασμένες εκτιμήσεις της περιεχόμενης ποσότητας των υδατανθράκων στα λαμβανόμενα γεύματα. Το δεύτερο μέρος της εργασίας αφορά στην ανάπτυξη μοντέλων εκτίμησης της πιθανότητας ατόμων με ΣΔ Τύπου Ι και Τύπου ΙΙ να εμφανίσουν σε βάθος χρόνου μακροπρόθεσμες επιπλοκές του ΣΔ, εστιάζοντας στη διαβητική αμφιβληστροειδοπάθεια. Για το σκοπό αυτό, εφαρμόστηκαν τεχνικές ταξινόμησης των δεδομένων, με χρήση τεχνητών νευρωνικών δικτύων με κυματιδιακές συναρτήσεις ενεργοποίησης. Για την ανάπτυξη και την αξιολόγηση των συστημάτων χρησιμοποιήθηκαν ιατρικά δεδομένα ατόμων με ΣΔ Τύπου Ι και Τύπου ΙΙ, που παραχωρήθηκαν από την Α’ Παιδιατρική Κλινική, Διαβητολογικό Κέντρο του Νοσοκομείου Π & Α Κυριακού, καθώς και από το Διαβητολογικό Κέντρο του Ιπποκράτειου Νοσοκομείου Αθηνών.

This study explores the ability of regression models, with no knowledge of the underlying physiology, to estimate physiological parameters relevant for metabolism and endocrinology. Four regression models were compared: multiple linear regression (MLR), principal component regression (PCR), partial least-squares regression (PLS) and regression using artificial neural networks (ANN). The pathway of mammalian gluconeogenesis was analyzed using [U−13C]glucose as tracer. A set of data was simulated by randomly selecting physiologically appropriate metabolic fluxes for the 9 steps of this pathway as independent variables. The isotope labeling patterns of key intermediates in the pathway were then calculated for each set of fluxes, yielding 29 dependent variables. Two thousand sets were created, allowing independent training and test data. Regression models were asked to predict the nine fluxes, given only the 29 isotopomers. For large training sets (>50) the artificial neural network model was superior, capturing 95% of the variability in the gluconeogenic flux, whereas the three linear models captured only 75%. This reflects the ability of neural networks to capture the inherent non-linearities of the metabolic system. The effect of error in the variables and the addition of random variables to the data set was considered. Model sensitivities were used to find the isotopomers that most influenced the predicted flux values. These studies provide the first test of multivariate regression models for the analysis of isotopomer flux data. They provide insight for metabolomics and the future of isotopic tracers in metabolic research where the underlying physiology is complex or unknown.