De novo prediction of RNA 3D structures with deep generative models.

We present a Deep Learning approach to predict 3D folding structures of RNAs from their nucleic acid sequence. Our approach combines an autoregressive Deep Generative Model, Monte Carlo Tree Search, and a score model to find and rank the most likely folding structures for a given RNA sequence. We show that RNA de novo structure prediction by deep learning is possible at atom resolution, despite the low number of experimentally measured structures that can be used for training. We confirm the predictive power of our approach by achieving competitive results in a retrospective evaluation of the RNA-Puzzles prediction challenges, without using structural contact information from multiple sequence alignments or additional data from chemical probing experiments. Blind predictions for recent RNA-Puzzle challenges under the name "Dfold" further support the competitive performance of our approach.

Wearable based monitoring and self-supervised contrastive learning detect clinical complications during treatment of Hematologic malignancies.

Serious clinical complications (SCC; CTCAE grade ≥ 3) occur frequently in patients treated for hematological malignancies. Early diagnosis and treatment of SCC are essential to improve outcomes. Here we report a deep learning model-derived SCC-Score to detect and predict SCC from time-series data recorded continuously by a medical wearable. In this single-arm, single-center, observational cohort study, vital signs and physical activity were recorded with a wearable for 31,234 h in 79 patients (54 Inpatient Cohort (IC)/25 Outpatient Cohort (OC)). Hours with normal physical functioning without evidence of SCC (regular hours) were presented to a deep neural network that was trained by a self-supervised contrastive learning objective to extract features from the time series that are typical in regular periods. The model was used to calculate a SCC-Score that measures the dissimilarity to regular features. Detection and prediction performance of the SCC-Score was compared to clinical documentation of SCC (AUROC ± SD). In total 124 clinically documented SCC occurred in the IC, 16 in the OC. Detection of SCC was achieved in the IC with a sensitivity of 79.7% and specificity of 87.9%, with AUROC of 0.91 ± 0.01 (OC sensitivity 77.4%, specificity 81.8%, AUROC 0.87 ± 0.02). Prediction of infectious SCC was possible up to 2 days before clinical diagnosis (AUROC 0.90 at -24 h and 0.88 at -48 h). We provide proof of principle for the detection and prediction of SCC in patients treated for hematological malignancies using wearable data and a deep learning model. As a consequence, remote patient monitoring may enable pre-emptive complication management.

Feasibility of Wearable-Based Remote Monitoring in Patients During Intensive Treatment for Aggressive Hematologic Malignancies.

PURPOSE: Intensive treatment protocols for aggressive hematologic malignancies harbor a high risk of serious clinical complications, such as infections. Current techniques of monitoring vital signs to detect such complications are cumbersome and often fail to diagnose them early. Continuous monitoring of vital signs and physical activity by means of an upper arm medical wearable allowing 24/7 streaming of such parameters may be a promising alternative. METHODS: This single-arm, single-center observational trial evaluated symptom-related patient-reported outcomes and feasibility of a wearable-based remote patient monitoring. All wearable data were reviewed retrospectively and were not available to the patient or clinical staff. A total of 79 patients (54 inpatients and 25 outpatients) participated and received standard-of-care treatment for a hematologic malignancy. In addition, the wearable was continuously worn and self-managed by the patient to record multiple parameters such as heart rate, oxygen saturation, and physical activity. RESULTS: Fifty-one patients (94.4%) in the inpatient cohort and 16 (64.0%) in the outpatient cohort reported gastrointestinal symptoms (diarrhea, nausea, and emesis), pain, dyspnea, or shivering in at least one visit. With the wearable, vital signs and physical activity were recorded for a total of 1,304.8 days. Recordings accounted for 78.0% (63.0-88.5; median [interquartile range]) of the potential recording time for the inpatient cohort and 84.6% (76.3-90.2) for the outpatient cohort. Adherence to the wearable was comparable in both cohorts, but decreased moderately over time during the trial. CONCLUSION: A high adherence to the wearable was observed in patients on intensive treatment protocols for a hematologic malignancy who experience high symptom burden. Remote patient monitoring of vital signs and physical activity was demonstrated to be feasible and of primarily sufficient quality.

Reliable Detection of Atrial Fibrillation with a Medical Wearable during Inpatient Conditions.

Atrial fibrillation (AF) is the most common arrhythmia and has a major impact on morbidity and mortality; however, detection of asymptomatic AF is challenging. This study sims to evaluate the sensitivity and specificity of non-invasive AF detection by a medical wearable. In this observational trial, patients with AF admitted to a hospital carried the wearable and an ECG Holter (control) in parallel over a period of 24 h, while not in a physically restricted condition. The wearable with a tight-fit upper armband employs a photoplethysmography technology to determine pulse rates and inter-beat intervals. Different algorithms (including a deep neural network) were applied to five-minute periods photoplethysmography datasets for the detection of AF. A total of 2306 h of parallel recording time could be obtained in 102 patients; 1781 h (77.2%) were automatically interpretable by an algorithm. Sensitivity to detect AF was 95.2% and specificity 92.5% (area under the receiver operating characteristics curve (AUC) 0.97). Usage of deep neural network improved the sensitivity of AF detection by 0.8% (96.0%) and specificity by 6.5% (99.0%) (AUC 0.98). Detection of AF by means of a wearable is feasible in hospitalized but physically active patients. Employing a deep neural network enables reliable and continuous monitoring of AF.

Information integration and decision making in flowering time control.

In order to successfully reproduce, plants must sense changes in their environment and flower at the correct time. Many plants utilize day length and vernalization, a mechanism for verifying that winter has occurred, to determine when to flower. Our study used available temperature and day length data from different climates to provide a general understanding how this information processing of environmental signals could have evolved in plants. For climates where temperature fluctuation correlations decayed exponentially, a simple stochastic model characterizing vernalization was able to reconstruct the switch-like behavior of the core flowering regulatory genes. For these and other climates, artificial neural networks were used to predict flowering gene expression patterns. For temperate plants, long-term cold temperature and short-term day length measurements were sufficient to produce robust flowering time decisions from the neural networks. Additionally, evolutionary simulations on neural networks confirmed that the combined signal of temperature and day length achieved the highest fitness relative to neural networks with access to only one of those inputs. We suggest that winter temperature memory is a well-adapted strategy for plants' detection of seasonal changes, and absolute day length is useful for the subsequent triggering of flowering.

OutCyte: a novel tool for predicting unconventional protein secretion.

The prediction of protein localization, such as in the extracellular space, from high-throughput data is essential for functional downstream inference. It is well accepted that some secreted proteins go through the classic endoplasmic reticulum-Golgi pathway with the guidance of a signal peptide. However, a large number of proteins have been found to reach the extracellular space by following unconventional secretory pathways. There remains a demand for reliable prediction of unconventional protein secretion (UPS). Here, we present OutCyte, a fast and accurate tool for the prediction of UPS, which for the first time has been built upon experimentally determined UPS proteins. OutCyte mediates the prediction of protein secretion in two steps: first, proteins with N-terminal signals are accurately filtered out; second, proteins without N-terminal signals are classified as UPS or intracellular proteins based on physicochemical features directly generated from their amino acid sequences. We are convinced that OutCyte will play a relevant role in the annotation of experimental data and will therefore contribute to further characterization of the extracellular nature of proteins by considering the commonly neglected UPS proteins.OutCyte has been implemented as a web server at www.outcyte.com.

Neural networks with circular filters enable data efficient inference of sequence motifs.

MOTIVATION: Nucleic acids and proteins often have localized sequence motifs that enable highly specific interactions. Due to the biological relevance of sequence motifs, numerous inference methods have been developed. Recently, convolutional neural networks (CNNs) have achieved state of the art performance. These methods were able to learn transcription factor binding sites from ChIP-seq data, resulting in accurate predictions on test data. However, CNNs typically distribute learned motifs across multiple filters, making them difficult to interpret. Furthermore, networks trained on small datasets often do not generalize well to new sequences. RESULTS: Here we present circular filters, a novel convolutional architecture, that convolves sequences with circularly permutated variants of the same filter. We motivate circular filters by the observation that CNNs frequently learn filters that correspond to shifted and truncated variants of the true motif. Circular filters enable learning of full-length motifs and allow easy interpretation of the learned filters. We show that circular filters improve motif inference performance over a wide range of hyperparameters as well as sequence length. Furthermore, we show that CNNs with circular filters in most cases outperform conventional CNNs at inferring DNA binding sites from ChIP-seq data. AVAILABILITY AND IMPLEMENTATION: Code is available at https://github.com/christopherblum. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Pharmacosimulation of interruptions and its solution in intravenous administration of cangrelor.

BACKGROUND: Cangrelor is an intravenous adenosine diphosphate (ADP) P2Y12 receptor antagonist, which has to be administered as a bolus followed by immediate infusion. Nevertheless, in clinical routine deviations from the correct practice, such as delayed infusion onset or interruptions during infusion, may occur. OBJECTIVE: The objective of the present study was to investigate the impact of administration delays on cangrelor concentration in a pharmacological simulation setting and to give possible solutions for the clinical practice. METHODS: We simulated the effects of different delays in administration of cangrelor in a model based on known pharmacokinetic parameters. Additionally, we calculated the optimal dosage of a second bolus. RESULTS: We demonstrate that already a short delay between the bolus and begin of infusion as well as short infusion interruptions considerably affect the serum concentration of cangrelor. Additionally, we estimate the dosage of a possible second bolus which highly depends on the duration of the delay. CONCLUSIONS: Our results emphasize that continuous administration of cangrelor is crucial to avoid the critical time frame of increased thrombosis risk. We suggest a strategy for dealing with interruptions by demonstrating that a second bolus allows to reach rapidly an effective but not excessive cangrelor serum concentration.

Automated computer-based detection of encounter behaviours in groups of honeybees.

Honeybees form societies in which thousands of members integrate their behaviours to act as a single functional unit. We have little knowledge on how the collaborative features are regulated by workers' activities because we lack methods that enable collection of simultaneous and continuous behavioural information for each worker bee. In this study, we introduce the Bee Behavioral Annotation System (BBAS), which enables the automated detection of bees' behaviours in small observation hives. Continuous information on position and orientation were obtained by marking worker bees with 2D barcodes in a small observation hive. We computed behavioural and social features from the tracking information to train a behaviour classifier for encounter behaviours (interaction of workers via antennation) using a machine learning-based system. The classifier correctly detected 93% of the encounter behaviours in a group of bees, whereas 13% of the falsely classified behaviours were unrelated to encounter behaviours. The possibility of building accurate classifiers for automatically annotating behaviours may allow for the examination of individual behaviours of worker bees in the social environments of small observation hives. We envisage that BBAS will be a powerful tool for detecting the effects of experimental manipulation of social attributes and sub-lethal effects of pesticides on behaviour.

Resource constrained flux balance analysis predicts selective pressure on the global structure of metabolic networks.

BACKGROUND: A universal feature of metabolic networks is their hourglass or bow-tie structure on cellular level. This architecture reflects the conversion of multiple input nutrients into multiple biomass components via a small set of precursor metabolites. However, it is yet unclear to what extent this structural feature is the result of natural selection. RESULTS: We extend flux balance analysis to account for limited cellular resources. Using this model, optimal structure of metabolic networks can be calculated for different environmental conditions. We observe a significant structural reshaping of metabolic networks for a toy-network and E. coli core metabolism if we increase the share of invested resources for switching between different nutrient conditions. Here, hub nodes emerge and the optimal network structure becomes bow-tie-like as a consequence of limited cellular resource constraint. We confirm this theoretical finding by comparing the reconstructed metabolic networks of bacterial species with respect to their lifestyle. CONCLUSIONS: We show that bow-tie structure can give a system-level fitness advantage to organisms that live in highly competitive and fluctuating environments. Here, limitation of cellular resources can lead to an efficiency-flexibility tradeoff where it pays off for the organism to shorten catabolic pathways if they are frequently activated and deactivated. As a consequence, generalists that shuttle between diverse environmental conditions should have a more predominant bow-tie structure than specialists that visit just a few isomorphic habitats during their life cycle.

Perception and regulatory principles of microbial growth control.

Fast growth represents an effective strategy for microbial organisms to survive in competitive environments. To accomplish this task, cells must adapt their metabolism to changing nutrient conditions in a way that maximizes their growth rate. However, the regulation of the growth related metabolic pathways can be fundamentally different among microbes. We therefore asked whether growth control by perception of the cell's intracellular metabolic state can give rise to higher growth than by direct perception of extracellular nutrient availability. To answer this question, we created a simplified dynamical computer model of a cellular metabolic network whose regulation was inferred by an optimization approach. We used this model for a competing species experiment, where a species with extracellular nutrient perception competes against one with intracellular nutrient perception by evaluating their respective average growth rate. We found that the intracellular perception is advantageous under situations where the up and down regulation of pathways cannot follow the fast changing nutrient availability in the environment. In this case, optimal regulation ignores any other nutrients except the most preferential ones, in agreement with the phenomenon of catabolite repression in prokaryotes. The corresponding metabolic pathways remain activated, despite environmental fluctuations. Therefore, the cell can take up preferential nutrients as soon as they are available without any prior regulation. As a result species that rely on intracellular perception gain a relevant fitness advantage in fluctuating nutrient environments, which enables survival by outgrowing competitors.

Exponential signaling gain at the receptor level enhances signal-to-noise ratio in bacterial chemotaxis.

Cellular signaling systems show astonishing precision in their response to external stimuli despite strong fluctuations in the molecular components that determine pathway activity. To control the effects of noise on signaling most efficiently, living cells employ compensatory mechanisms that reach from simple negative feedback loops to robustly designed signaling architectures. Here, we report on a novel control mechanism that allows living cells to keep precision in their signaling characteristics - stationary pathway output, response amplitude, and relaxation time - in the presence of strong intracellular perturbations. The concept relies on the surprising fact that for systems showing perfect adaptation an exponential signal amplification at the receptor level suffices to eliminate slowly varying multiplicative noise. To show this mechanism at work in living systems, we quantified the response dynamics of the E. coli chemotaxis network after genetically perturbing the information flux between upstream and downstream signaling components. We give strong evidence that this signaling system results in dynamic invariance of the activated response regulator against multiplicative intracellular noise. We further demonstrate that for environmental conditions, for which precision in chemosensing is crucial, the invariant response behavior results in highest chemotactic efficiency. Our results resolve several puzzling features of the chemotaxis pathway that are widely conserved across prokaryotes but so far could not be attributed any functional role.

Thermal robustness of signaling in bacterial chemotaxis.

Temperature is a global factor that affects the performance of all intracellular networks. Robustness against temperature variations is thus expected to be an essential network property, particularly in organisms without inherent temperature control. Here, we combine experimental analyses with computational modeling to investigate thermal robustness of signaling in chemotaxis of Escherichia coli, a relatively simple and well-established model for systems biology. We show that steady-state and kinetic pathway parameters that are essential for chemotactic performance are indeed temperature-compensated in the entire physiological range. Thermal robustness of steady-state pathway output is ensured at several levels by mutual compensation of temperature effects on activities of individual pathway components. Moreover, the effect of temperature on adaptation kinetics is counterbalanced by preprogrammed temperature dependence of enzyme synthesis and stability to achieve nearly optimal performance at the growth temperature. Similar compensatory mechanisms are expected to ensure thermal robustness in other systems.

Robust signal processing in living cells.

Cellular signaling networks have evolved an astonishing ability to function reliably and with high fidelity in uncertain environments. A crucial prerequisite for the high precision exhibited by many signaling circuits is their ability to keep the concentrations of active signaling compounds within tightly defined bounds, despite strong stochastic fluctuations in copy numbers and other detrimental influences. Based on a simple mathematical formalism, we identify topological organizing principles that facilitate such robust control of intracellular concentrations in the face of multifarious perturbations. Our framework allows us to judge whether a multiple-input-multiple-output reaction network is robust against large perturbations of network parameters and enables the predictive design of perfectly robust synthetic network architectures. Utilizing the Escherichia coli chemotaxis pathway as a hallmark example, we provide experimental evidence that our framework indeed allows us to unravel the topological organization of robust signaling. We demonstrate that the specific organization of the pathway allows the system to maintain global concentration robustness of the diffusible response regulator CheY with respect to several dominant perturbations. Our framework provides a counterpoint to the hypothesis that cellular function relies on an extensive machinery to fine-tune or control intracellular parameters. Rather, we suggest that for a large class of perturbations, there exists an appropriate topology that renders the network output invariant to the respective perturbations.

A sequestration feedback determines dynamics and temperature entrainment of the KaiABC circadian clock.

The circadian rhythm of the cyanobacterium Synechococcus elongatus is controlled by three proteins, KaiA, KaiB, and KaiC. In a test tube, these proteins form complexes of various stoichiometry and the average phosphorylation level of KaiC exhibits robust circadian oscillations in the presence of ATP. Using mathematical modeling, we were able to reproduce quantitatively the experimentally observed phosphorylation dynamics of the KaiABC clockwork in vitro. We thereby identified a highly non-linear feedback loop through KaiA inactivation as the key synchronization mechanism of KaiC phosphorylation. By using the novel method of native mass spectrometry, we confirm the theoretically predicted complex formation dynamics and show that inactivation of KaiA is a consequence of sequestration by KaiC hexamers and KaiBC complexes. To test further the predictive power of the mathematical model, we reproduced the observed phase synchronization dynamics on entrainment by temperature cycles. Our model gives strong evidence that the underlying entrainment mechanism arises from a temperature-dependent change in the abundance of KaiAC and KaiBC complexes.

Role of translational coupling in robustness of bacterial chemotaxis pathway.

Chemotaxis allows bacteria to colonize their environment more efficiently and to find optimal growth conditions, and is consequently under strong evolutionary selection. Theoretical and experimental analyses of bacterial chemotaxis suggested that the pathway has been evolutionarily optimized to produce robust output under conditions of such physiological perturbations as stochastic intercellular variations in protein levels while at the same time minimizing complexity and cost of protein expression. Pathway topology in Escherichia coli apparently evolved to produce an invariant output under concerted variations in protein levels, consistent with experimentally observed transcriptional coupling of chemotaxis genes. Here, we show that the pathway robustness is further enhanced through the pairwise translational coupling of adjacent genes. Computer simulations predicted that the robustness of the pathway against the uncorrelated variations in protein levels can be enhanced by a selective pairwise coupling of individual chemotaxis genes on one mRNA, with the order of genes in E. coli ranking among the best in terms of noise compensation. Translational coupling between chemotaxis genes was experimentally confirmed, and coupled expression of these genes was shown to improve chemotaxis. Bioinformatics analysis further revealed that E. coli gene order corresponds to consensus in sequenced bacterial genomes, confirming evolutionary selection for noise reduction. Since polycistronic gene organization is common in bacteria, translational coupling between adjacent genes may provide a general mechanism to enhance robustness of their signaling and metabolic networks. Moreover, coupling between expression of neighboring genes is also present in eukaryotes, and similar principles of noise reduction might thus apply to all cellular networks.

Signatures of gene expression noise in cellular systems.

Noise in gene expression, either due to inherent stochasticity or to varying inter- and intracellular environment, can generate significant cell-to-cell variability of protein levels in clonal populations. To quantify the different sources of gene expression noise, several theoretical studies have been performed using either a quasi-stationary approximation for the emerging master equation or employing a time-dependent description, when cell division is taken explicitly into account. Here, we give an overview of the different origins of gene expression noise which were found experimentally and introduce the basic stochastic modeling approaches. We extend, and apply a time-dependent description of gene expression noise to experimental data. The analysis shows that the induction level of the transcription factor can be employed to discriminate the noise profiles and their characteristic signatures. On the basis of experimentally measured cell distributions, our simulations suggest that transcription factor binding and promoter activation can be modeled independently of each other with sufficient accuracy.

Quantifying origins of cell-to-cell variations in gene expression.

A general dynamic description of protein synthesis was employed to quantify different sources of gene expression noise in cellular systems. To test our approach, we use time-resolved expression data of individual human cells and, from this information, predict the stationary cell-to-cell variation in protein levels in a clonal population. For three of the four human genes investigated, the cellular variations in expression level are not due to fluctuations in promoter activity or transcript copy number, but are almost exclusively a consequence of long-term variations of gene regulatory factors or the global cellular state. Moreover, we show that a dynamic description is much more reliable to discriminate extrinsic and intrinsic sources of noise than it is on grounds of cell-cycle averaged descriptions. The excellent agreement between the theoretical predictions and the experimentally measured noise strengths shows that a quantitative description of gene expression noise is indeed possible on the basis of idealized stochastic processes.

Functioning and robustness of a bacterial circadian clock.

Cyanobacteria are the simplest known cellular systems that regulate their biological activities in daily cycles. For the cyanobacterium Synechococcus elongatus, it has been shown by in vitro and in vivo experiments that the basic circadian timing process is based on rhythmic phosphorylation of KaiC hexamers. Despite the excellent experimental work, a full systems level understanding of the in vitro clock is still lacking. In this work, we provide a mathematical approach to scan different hypothetical mechanisms for the primary circadian oscillator, starting from experimentally established molecular properties of the clock proteins. Although optimised for highest performance, only one of the in silico-generated reaction networks was able to reproduce the experimentally found high amplitude and robustness against perturbations. In this reaction network, a negative feedback synchronises the phosphorylation level of the individual hexamers and has indeed been realised in S. elongatus by KaiA sequestration as confirmed by experiments.

Co-expression of signaling proteins improves robustness of the bacterial chemotaxis pathway.

Biological systems are exposed to various perturbations that affect performance of the cellular networks, with stochastic variation in protein levels, or gene expression noise, being one of the major sources of intracellular perturbations. We recently used Escherichia coli chemotaxis as a model to analyze robustness against such noise and demonstrated theoretically and experimentally that a steady-state output of the pathway is robust against concerted variation in the levels of all chemotaxis proteins. However, our model predicted that the pathway topology does not confer much robustness against an uncorrelated variation in the protein levels. To test whether additional robustness features might be missing from our model, we compare here its predictions with an experimentally determined chemotactic performance under varying levels of individual proteins. Our data show that the pathway is indeed even more robust than predicted to two types of perturbations-the variation in the levels of the adaptation enzymes and a correlated expression of CheY and CheZ. Although the design features that are responsible for this higher robustness still remain to be understood, our results stress the importance of a robust design of both native and synthetic signaling networks.