General value functions for fault detection in multivariate time series data.

One of the greatest challenges to the automated production of goods is equipment malfunction. Ideally, machines should be able to automatically predict and detect operational faults in order to minimize downtime and plan for timely maintenance. While traditional condition-based maintenance (CBM) involves costly sensor additions and engineering, machine learning approaches offer the potential to learn from already existing sensors. Implementations of data-driven CBM typically use supervised and semi-supervised learning to classify faults. In addition to a large collection of operation data, records of faulty operation are also necessary, which are often costly to obtain. Instead of classifying faults, we use an approach to detect abnormal behaviour within the machine's operation. This approach is analogous to semi-supervised anomaly detection in machine learning (ML), with important distinctions in experimental design and evaluation specific to the problem of industrial fault detection. We present a novel method of machine fault detection using temporal-difference learning and General Value Functions (GVFs). Using GVFs, we form a predictive model of sensor data to detect faulty behaviour. As sensor data from machines is not i.i.d. but closer to Markovian sampling, temporal-difference learning methods should be well suited for this data. We compare our GVF outlier detection (GVFOD) algorithm to a broad selection of multivariate and temporal outlier detection methods, using datasets collected from a tabletop robot emulating the movement of an industrial actuator. We find that not only does GVFOD achieve the same recall score as other multivariate OD algorithms, it attains significantly higher precision. Furthermore, GVFOD has intuitive hyperparameters which can be selected based upon expert knowledge of the application. Together, these findings allow for a more reliable detection of abnormal machine behaviour to allow ideal timing of maintenance; saving resources, time and cost.

A shift towards succinate-producing Prevotella in the ruminal microbiome challenged with monensin.

The time-resolved impact of monensin on the active rumen microbiome was studied in a rumen-simulating technique (Rusitec) with metaproteomic and metabolomic approaches. Monensin treatment caused a decreased fibre degradation potential that was observed by the reduced abundance of proteins assigned to fibrolytic bacteria and glycoside hydrolases, sugar transporters and carbohydrate metabolism. Decreased proteolytic activities resulted in reduced amounts of ammonium as well as branched-chain fatty acids. The family Prevotellaceae exhibited increased resilience in the presence of monensin, with a switch of the metabolism from acetate to succinate production. Prevotella species harbour a membrane-bound electron transfer complex, which drives the reduction of fumarate to succinate, which is the substrate for propionate production in the rumen habitat. Besides the increased succinate production, a concomitant depletion of methane concentration was observed upon monensin exposure. Our study demonstrates that Prevotella sp. shifts its metabolism successfully in response to monensin exposure and Prevotellaceae represents the key bacterial family stabilizing the rumen microbiota during exposure to monensin.

ANARI: A 3-D Rendering API Standard.

ANARI is a new 3-D rendering API, an emerging Khronos standard that enables visualization applications to leverage the state-of-the-art rendering techniques across diverse hardware platforms and rendering engines. Visualization applications have historically embedded custom-written renderers to enable them to provide the necessary combination of features, performance, and visual fidelity required by their users. As computing power, rendering algorithms, dedicated rendering hardware acceleration operations, and associated low-level APIs have advanced, the effort and costs associated with maintaining renderers within visualization applications have risen dramatically. The rising cost and complexity associated with renderer development creates an undesirable barrier for visualization applications to be able to fully benefit from the latest rendering methods and hardware. ANARI directly addresses these challenges by providing a high-level, visualization-oriented API that abstracts low-level rendering algorithms and hardware acceleration details while providing easy and efficient access to diverse ANARI implementations, thereby enabling visualization applications to support the state-of-the-art rendering capabilities.

Composite Recurrent Convolutional Neural Networks Offer a Position-Aware Prosthesis Control Alternative While Balancing Predictive Accuracy with Training Burden.

To mitigate the "limb position effect" that hinders myoelectric upper limb prosthesis control, pattern recognition-based models must accurately predict user-intended movements across a multitude of limb positions. Such models can use electromyography (EMG) and inertial measurement units to capture necessary multi-position data. However, this data capture solution requires lengthy user-performed model training routines, with movements in many limb positions, plus retraining thereafter due to inherent signal variations over time. While a general-purpose control model (trained with a dataset that represents numerous device users) eliminates the user-training requirement altogether, it yields low movement predictive accuracy. Conversely, a user-specific control model (trained with a smaller dataset from an individual) yields high predictive accuracy, but requires retraining over time. This study capitalizes on the benefits offered by both such control options, and contributes an alternative control solution-a novel recurrent convolutional neural network (RCNN)-based Composite Model that combines the representation portion of a general-purpose model, with the decision portion of a user-specific model. The resulting Composite Model offers moderate movement predictive accuracy across various limb positions and a reduction in user training routine requirements, suggesting a new research direction to help mitigate the limb position effect along with model training burden.

Prediction, Knowledge, and Explainability: Examining the Use of General Value Functions in Machine Knowledge.

Within computational reinforcement learning, a growing body of work seeks to express an agent's knowledge of its world through large collections of predictions. While systems that encode predictions as General Value Functions (GVFs) have seen numerous developments in both theory and application, whether such approaches are explainable is unexplored. In this perspective piece, we explore GVFs as a form of explainable AI. To do so, we articulate a subjective agent-centric approach to explainability in sequential decision-making tasks. We propose that prior to explaining its decisions to others, an self-supervised agent must be able to introspectively explain decisions to itself. To clarify this point, we review prior applications of GVFs that involve human-agent collaboration. In doing so, we demonstrate that by making their subjective explanations public, predictive knowledge agents can improve the clarity of their operation in collaborative tasks.

Structural basis of DNA methylation-dependent site selectivity of the Epstein-Barr virus lytic switch protein ZEBRA/Zta/BZLF1.

In infected cells, Epstein-Barr virus (EBV) alternates between latency and lytic replication. The viral bZIP transcription factor ZEBRA (Zta, BZLF1) regulates this cycle by binding to two classes of ZEBRA response elements (ZREs): CpG-free motifs resembling the consensus AP-1 site recognized by cellular bZIP proteins and CpG-containing motifs that are selectively bound by ZEBRA upon cytosine methylation. We report structural and mutational analysis of ZEBRA bound to a CpG-methylated ZRE (meZRE) from a viral lytic promoter. ZEBRA recognizes the CpG methylation marks through a ZEBRA-specific serine and a methylcytosine-arginine-guanine triad resembling that found in canonical methyl-CpG binding proteins. ZEBRA preferentially binds the meZRE over the AP-1 site but mutating the ZEBRA-specific serine to alanine inverts this selectivity and abrogates viral replication. Our findings elucidate a DNA methylation-dependent switch in ZEBRA's transactivation function that enables ZEBRA to bind AP-1 sites and promote viral latency early during infection and subsequently, under appropriate conditions, to trigger EBV lytic replication by binding meZREs.

Cyclic Crack Growth in Chemically Tailored Isotropic Austenitic Steel Processed by Electron Beam Powder Bed Fusion.

The present study analyzes the cyclic crack propagation behavior in an austenitic steel processed by electron beam powder bed fusion (PBF-EB). The threshold value of crack growth as well as the crack growth behavior in the Paris regime were studied. In contrast to other austenitic steels, the building direction during PBF-EB did not affect the crack propagation rate, i.e., the crack growth rates perpendicular and parallel to the building direction were similar due to the isotropic microstructure characterized by equiaxed grains. Furthermore, the influence of significantly different building parameters was studied and, thereby, different energy inputs causing locally varying manganese content. Crack growth behavior was not affected by these changes. Even a compositional gradation within the same specimen, i.e., crack growth through an interface of areas with high and areas with low manganese content, did not lead to a significant change of the crack growth rate. Thus, the steel studied is characterized by a quite robust cyclic crack growth behavior independent from building direction and hardly affected by typical parameter deviations in the PBF-EB process.

Interpretable PID parameter tuning for control engineering using general dynamic neural networks: An extensive comparison.

Modern automation systems largely rely on closed loop control, wherein a controller interacts with a controlled process via actions, based on observations. These systems are increasingly complex, yet most deployed controllers are linear Proportional-Integral-Derivative (PID) controllers. PID controllers perform well on linear and near-linear systems but their simplicity is at odds with the robustness required to reliably control complex processes. Modern machine learning techniques offer a way to extend PID controllers beyond their linear control capabilities by using neural networks. However, such an extension comes at the cost of losing stability guarantees and controller interpretability. In this paper, we examine the utility of extending PID controllers with recurrent neural networks--namely, General Dynamic Neural Networks (GDNN); we show that GDNN (neural) PID controllers perform well on a range of complex control systems and highlight how they can be a scalable and interpretable option for modern control systems. To do so, we provide an extensive study using four benchmark systems that represent the most common control engineering benchmarks. All control environments are evaluated with and without noise as well as with and without disturbances. The neural PID controller performs better than standard PID control in 15 of 16 tasks and better than model-based control in 13 of 16 tasks. As a second contribution, we address the lack of interpretability that prevents neural networks from being used in real-world control processes. We use bounded-input bounded-output stability analysis to evaluate the parameters suggested by the neural network, making them understandable for engineers. This combination of rigorous evaluation paired with better interpretability is an important step towards the acceptance of neural-network-based control approaches for real-world systems. It is furthermore an important step towards interpretable and safely applied artificial intelligence.

Examining the Use of Temporal-Difference Incremental Delta-Bar-Delta for Real-World Predictive Knowledge Architectures.

Predictions and predictive knowledge have seen recent success in improving not only robot control but also other applications ranging from industrial process control to rehabilitation. A property that makes these predictive approaches well-suited for robotics is that they can be learned online and incrementally through interaction with the environment. However, a remaining challenge for many prediction-learning approaches is an appropriate choice of prediction-learning parameters, especially parameters that control the magnitude of a learning machine's updates to its predictions (the learning rates or step sizes). Typically, these parameters are chosen based on an extensive parameter search-an approach that neither scales well nor is well-suited for tasks that require changing step sizes due to non-stationarity. To begin to address this challenge, we examine the use of online step-size adaptation using the Modular Prosthetic Limb: a sensor-rich robotic arm intended for use by persons with amputations. Our method of choice, Temporal-Difference Incremental Delta-Bar-Delta (TIDBD), learns and adapts step sizes on a feature level; importantly, TIDBD allows step-size tuning and representation learning to occur at the same time. As a first contribution, we show that TIDBD is a practical alternative for classic Temporal-Difference (TD) learning via an extensive parameter search. Both approaches perform comparably in terms of predicting future aspects of a robotic data stream, but TD only achieves comparable performance with a carefully hand-tuned learning rate, while TIDBD uses a robust meta-parameter and tunes its own learning rates. Secondly, our results show that for this particular application TIDBD allows the system to automatically detect patterns characteristic of sensor failures common to a number of robotic applications. As a third contribution, we investigate the sensitivity of classic TD and TIDBD with respect to the initial step-size values on our robotic data set, reaffirming the robustness of TIDBD as shown in previous papers. Together, these results promise to improve the ability of robotic devices to learn from interactions with their environments in a robust way, providing key capabilities for autonomous agents and robots.

Enzymatic generation of lactulose in sweet and acid whey: Optimization of feed composition and structural elucidation of 1-lactulose.

Prebiotics are rising in interest in commercial scale productions due to increasing health awareness of consumers. Under bio-economic aspects, sweet and acid whey provide a suitable feed medium for the enzymatic generation of prebiotic lactulose. Since whey has a broad variation in composition, the influence of the feed composition on the concentration of generated lactulose was investigated. The influence of lactose and fructose concentration as well as enzymatic activity of two commercially available ß-galactosidases were investigated. The results were evaluated via response surface analysis with a quadratic model containing pairwise interaction terms. The optimal feed composition yielding a theoretical maximal amount of lactulose was determined as 1.28 or 0.74 mol/kg fructose and 0.17 or 0.19 mol/kg lactose with an enzymatic activity of 2.0 or 2.8 µkat/kg for acid (pH 4.4) or sweet (pH 6.6) whey. Furthermore, the major reaction product was isolated and subsequently, the structural identity was elucidated and verified via extensive NMR analysis.

Direct 1H NMR Quantitation of Valuable Furan Fatty Acids in Fish Oils and Fish Oil Fractions.

Furan fatty acids (FuFAs) are a class of naturally occurring minor fatty acids with fish as the richest food source. Typically, FuFA analysis is cumbersome and involves several steps. We developed a quantitative 1H NMR method (qNMR) in which fish oil samples were directly measured after dilution with CDCl3 stabilized with silver (which was essential to prevent formation of radicals) and addition of an internal standard. The singlet at δ = 1.89 ppm was suitable for quantitation of monomethyl FuFAs, whereas the signal at δ = 1.83 ppm was suitable to quantitate dimethyl FuFAs. Using standard NMR tubes with 650 µL solvent, the limit of quantitation was 0.5 µg (dimethyl FuFAs) and 1.0 µg (monomethyl FuFAs). Applied to three fish oil and two enriched fish oil samples (sample weight, 10 mg), the final qNMR method resulted in similar total FuFA contents as determined in parallel by gas chromatography coupled to mass spectrometry.

Correction to: Novel metagenome-derived ornithine lipids identified by functional screening for biosurfactants.

The published online version contains mistake in the author list.

Staufen2-mediated RNA recognition and localization requires combinatorial action of multiple domains.

Throughout metazoans, Staufen (Stau) proteins are core factors of mRNA localization particles. They consist of three to four double-stranded RNA binding domains (dsRBDs) and a C-terminal dsRBD-like domain. Mouse Staufen2 (mStau2)-like Drosophila Stau (dmStau) contains four dsRBDs. Existing data suggest that only dsRBDs 3-4 are necessary and sufficient for mRNA binding. Here, we show that dsRBDs 1 and 2 of mStau2 bind RNA with similar affinities and kinetics as dsRBDs 3 and 4. While RNA binding by these tandem domains is transient, all four dsRBDs recognize their target RNAs with high stability. Rescue experiments in Drosophila oocytes demonstrate that mStau2 partially rescues dmStau-dependent mRNA localization. In contrast, a rescue with mStau2 bearing RNA-binding mutations in dsRBD1-2 fails, confirming the physiological relevance of our findings. In summary, our data show that the dsRBDs 1-2 play essential roles in the mRNA recognition and function of Stau-family proteins of different species.

Ultrashort Broadband Cooperative Pulses for Multidimensional Biomolecular NMR Experiments.

NMR spectroscopy at ultra-high magnetic fields requires improved radiofrequency (rf) pulses to cover the increased spectral bandwidth. Optimized 90° pulse pairs were introduced as Ramsey-type cooperative (Ram-COOP) pulses for biomolecular NMR applications. The Ram-COOP element provides broadband excitation with enhanced sensitivity and reduced artifacts even at magnetic fields >1.0 GHz 1 H Larmor frequency (23 T). A pair of 30 µs Ram-COOP pulses achieves an excitation bandwidth of 100 kHz with a maximum rf field of 20 kHz, more than three-fold improved compared to excitation by rectangular pulses. Ram-COOP pulses exhibit little offset-dependent phase errors and are robust to rf inhomogeneity. The performance of the Ram-COOP element is experimentally confirmed with heteronuclear multidimensional NMR experiments, applied to proteins and nucleic acids. Ram-COOP provides broadband excitation at low rf field strength suitable for application at current magnetic fields and beyond 23 T.

RNA structure refinement using NMR solvent accessibility data.

NMR spectroscopy is a powerful technique to study ribonucleic acids (RNAs) which are key players in a plethora of cellular processes. Although the NMR toolbox for structural studies of RNAs expanded during the last decades, they often remain challenging. Here, we show that solvent paramagnetic relaxation enhancements (sPRE) induced by the soluble, paramagnetic compound Gd(DTPA-BMA) provide a quantitative measure for RNA solvent accessibility and encode distance-to-surface information that correlates well with RNA structure and improves accuracy and convergence of RNA structure determination. Moreover, we show that sPRE data can be easily obtained for RNAs with any isotope labeling scheme and is advantageous regarding sample preparation, stability and recovery. sPRE data show a large dynamic range and reflect the global fold of the RNA suggesting that they are well suited to identify interaction surfaces, to score structural models and as restraints in RNA structure determination.

Synchronization and Propagation of Global Sleep Spindles.

Sleep spindles occur thousands of times during normal sleep and can be easily detected by visual inspection of EEG signals. These characteristics make spindles one of the most studied EEG structures in mammalian sleep. In this work we considered global spindles, which are spindles that are observed simultaneously in all EEG channels. We propose a methodology that investigates both the signal envelope and phase/frequency of each global spindle. By analysing the global spindle phase we showed that 90% of spindles synchronize with an average latency time of 0.1 s. We also measured the frequency modulation (chirp) of global spindles and found that global spindle chirp and synchronization are not correlated. By investigating the signal envelopes and implementing a homogeneous and isotropic propagation model, we could estimate both the signal origin and velocity in global spindles. Our results indicate that this simple and non-invasive approach could determine with reasonable precision the spindle origin, and allowed us to estimate a signal speed of 0.12 m/s. Finally, we consider whether synchronization might be useful as a non-invasive diagnostic tool.

Identification of two forms of TNF tolerance in human monocytes: differential inhibition of NF-κB/AP-1- and PP1-associated signaling.

The molecular basis of TNF tolerance is poorly understood. In human monocytes we detected two forms of TNF refractoriness, as follows: absolute tolerance was selective, dose dependently affecting a small group of powerful effector molecules; induction tolerance represented a more general phenomenon. Preincubation with a high TNF dose induces both absolute and induction tolerance, whereas low-dose preincubation predominantly mediates absolute tolerance. In cells preincubated with the high TNF dose, we observed blockade of IκBα phosphorylation/proteolysis and nuclear p65 translocation. More prominent in cells preincubated with the high dose, reduced basal IκBα levels were found, accompanied by increased IκBα degradation, suggesting an increased IκBα turnover. In addition, a nuclear elevation of p50 was detected in tolerant cells, which was more visible following high-dose preincubation. TNF-induced phosphorylation of p65-Ser(536), p38, and c-jun was inhibited, and basal inhibitory p65-Ser(468) phosphorylation was increased in tolerant cells. TNF tolerance induced by the low preincubation dose is mediated by glycogen synthesis kinase-3, whereas high-dose preincubation-mediated tolerance is regulated by A20/glycogen synthesis kinase-3 and protein phosphatase 1-dependent mechanisms. To our knowledge, we present the first genome-wide analysis of TNF tolerance in monocytic cells, which differentially inhibits NF-κB/AP-1-associated signaling and shifts the kinase/phosphatase balance. These forms of refractoriness may provide a cellular paradigm for resolution of inflammation and may be involved in immune paralysis.

Regulation of monocyte differentiation by specific signaling modules and associated transcription factor networks.

Monocyte/macrophages are important players in orchestrating the immune response as well as connecting innate and adaptive immunity. Myelopoiesis and monopoiesis are characterized by the interplay between expansion of stem/progenitor cells and progression towards further developed (myelo)monocytic phenotypes. In response to a variety of differentiation-inducing stimuli, various prominent signaling pathways are activated. Subsequently, specific transcription factors are induced, regulating cell proliferation and maturation. This review article focuses on the integration of signaling modules and transcriptional networks involved in the determination of monocytic differentiation.