Response to Comment on "Pushing the frontiers of density functionals by solving the fractional electron problem".

Gerasimov et al. claim that the ability of DM21 to respect fractional charge (FC) and fractional spin (FS) conditions outside of the training set has not been demonstrated in our paper. This is based on (i) asserting that the training set has a ~50% overlap with our bond-breaking benchmark (BBB) and (ii) questioning the validity and accuracy of our other generalization examples. We disagree with their analysis and believe that the points raised are either incorrect or not relevant to the main conclusions of the paper and to the assessment of general quality of DM21.

[Corrosion damage in oesophagus after digesting a button battery].

Button batteries are present in most households, i.e. in toys, hearing aids and remote controls. Due to technical progression button batteries have become increasingly powerful and have simultaneously increased the risk of severe complications when ingested. In this case report, an X-ray of a ten-month-old baby revealed a button battery trapped in the upper part of oesophagus. The battery was removed within two hours from the time of swallowing, but the battery had inflicted severe damage of the oesophageal mouth. The patient was hospitalized for three days, controlled for three months and showed no signs of swallowing difficulties or other sequelae.

AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models.

The AlphaFold Protein Structure Database (AlphaFold DB, https://alphafold.ebi.ac.uk) is an openly accessible, extensive database of high-accuracy protein-structure predictions. Powered by AlphaFold v2.0 of DeepMind, it has enabled an unprecedented expansion of the structural coverage of the known protein-sequence space. AlphaFold DB provides programmatic access to and interactive visualization of predicted atomic coordinates, per-residue and pairwise model-confidence estimates and predicted aligned errors. The initial release of AlphaFold DB contains over 360,000 predicted structures across 21 model-organism proteomes, which will soon be expanded to cover most of the (over 100 million) representative sequences from the UniRef90 data set.

Pushing the frontiers of density functionals by solving the fractional electron problem.

Density functional theory describes matter at the quantum level, but all popular approximations suffer from systematic errors that arise from the violation of mathematical properties of the exact functional. We overcame this fundamental limitation by training a neural network on molecular data and on fictitious systems with fractional charge and spin. The resulting functional, DM21 (DeepMind 21), correctly describes typical examples of artificial charge delocalization and strong correlation and performs better than traditional functionals on thorough benchmarks for main-group atoms and molecules. DM21 accurately models complex systems such as hydrogen chains, charged DNA base pairs, and diradical transition states. More crucially for the field, because our methodology relies on data and constraints, which are continually improving, it represents a viable pathway toward the exact universal functional.

Applying and improving AlphaFold at CASP14.

We describe the operation and improvement of AlphaFold, the system that was entered by the team AlphaFold2 to the "human" category in the 14th Critical Assessment of Protein Structure Prediction (CASP14). The AlphaFold system entered in CASP14 is entirely different to the one entered in CASP13. It used a novel end-to-end deep neural network trained to produce protein structures from amino acid sequence, multiple sequence alignments, and homologous proteins. In the assessors' ranking by summed z scores (>2.0), AlphaFold scored 244.0 compared to 90.8 by the next best group. The predictions made by AlphaFold had a median domain GDT_TS of 92.4; this is the first time that this level of average accuracy has been achieved during CASP, especially on the more difficult Free Modeling targets, and represents a significant improvement in the state of the art in protein structure prediction. We reported how AlphaFold was run as a human team during CASP14 and improved such that it now achieves an equivalent level of performance without intervention, opening the door to highly accurate large-scale structure prediction.

Autonomous Surface and Underwater Vehicles as Effective Ecosystem Monitoring and Research Platforms in the Arctic-The Glider Project.

Effective ocean management requires integrated and sustainable ocean observing systems enabling us to map and understand ecosystem properties and the effects of human activities. Autonomous subsurface and surface vehicles, here collectively referred to as "gliders", are part of such ocean observing systems providing high spatiotemporal resolution. In this paper, we present some of the results achieved through the project "Unmanned ocean vehicles, a flexible and cost-efficient offshore monitoring and data management approach-GLIDER". In this project, three autonomous surface and underwater vehicles were deployed along the Lofoten-Vesterålen (LoVe) shelf-slope-oceanic system, in Arctic Norway. The aim of this effort was to test whether gliders equipped with novel sensors could effectively perform ecosystem surveys by recording physical, biogeochemical, and biological data simultaneously. From March to September 2018, a period of high biological activity in the area, the gliders were able to record a set of environmental parameters, including temperature, salinity, and oxygen, map the spatiotemporal distribution of zooplankton, and record cetacean vocalizations and anthropogenic noise. A subset of these parameters was effectively employed in near-real-time data assimilative ocean circulation models, improving their local predictive skills. The results presented here demonstrate that autonomous gliders can be effective long-term, remote, noninvasive ecosystem monitoring and research platforms capable of operating in high-latitude marine ecosystems. Accordingly, these platforms can record high-quality baseline environmental data in areas where extractive activities are planned and provide much-needed information for operational and management purposes.

Highly accurate protein structure prediction with AlphaFold.

Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort1-4, the structures of around 100,000 unique proteins have been determined5, but this represents a small fraction of the billions of known protein sequences6,7. Structural coverage is bottlenecked by the months to years of painstaking effort required to determine a single protein structure. Accurate computational approaches are needed to address this gap and to enable large-scale structural bioinformatics. Predicting the three-dimensional structure that a protein will adopt based solely on its amino acid sequence-the structure prediction component of the 'protein folding problem'8-has been an important open research problem for more than 50 years9. Despite recent progress10-14, existing methods fall far short of atomic accuracy, especially when no homologous structure is available. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known. We validated an entirely redesigned version of our neural network-based model, AlphaFold, in the challenging 14th Critical Assessment of protein Structure Prediction (CASP14)15, demonstrating accuracy competitive with experimental structures in a majority of cases and greatly outperforming other methods. Underpinning the latest version of AlphaFold is a novel machine learning approach that incorporates physical and biological knowledge about protein structure, leveraging multi-sequence alignments, into the design of the deep learning algorithm.

Highly accurate protein structure prediction for the human proteome.

Protein structures can provide invaluable information, both for reasoning about biological processes and for enabling interventions such as structure-based drug development or targeted mutagenesis. After decades of effort, 17% of the total residues in human protein sequences are covered by an experimentally determined structure1. Here we markedly expand the structural coverage of the proteome by applying the state-of-the-art machine learning method, AlphaFold2, at a scale that covers almost the entire human proteome (98.5% of human proteins). The resulting dataset covers 58% of residues with a confident prediction, of which a subset (36% of all residues) have very high confidence. We introduce several metrics developed by building on the AlphaFold model and use them to interpret the dataset, identifying strong multi-domain predictions as well as regions that are likely to be disordered. Finally, we provide some case studies to illustrate how high-quality predictions could be used to generate biological hypotheses. We are making our predictions freely available to the community and anticipate that routine large-scale and high-accuracy structure prediction will become an important tool that will allow new questions to be addressed from a structural perspective.

[Treatment of patients with obstructive sleep apnoea syndrome].

Continuous positive airway pressure (CPAP) is an effective treatment modality for patients with obstructive sleep apnoea syndrome (OSAS). Surgical treatment of OSAS can include functional nasal surgery, uvulopalatopharyngoplasty, transoral robotic surgery, maxillo-mandibular advancement (MMA) and bariatric surgery. MMA should be considered in patients with moderate to severe OSAS, if CPAP treatment is ineffective or not tolerated, as well in patients with failure of previous sleep surgery or in patients with severe dentofacial anomalies. In this review, we stress, that multidisciplinary management between sleep medicine clinicians and surgeons is crucial.

Tiny but Fatty: Lipids and Fatty Acids in the Daubed Shanny (Leptoclinus Maculatus), a Small Fish in Svalbard Waters.

The seasonal dynamic of lipids and their fatty acid constituents in the lipid sac and muscles of pelagic postlarval Leptoclinus maculatus, an ecologically important fish species in the Arctic food nets, in Kongsfjord, Svalbard waters was studied. The determination of the qualitative and quantitative content of the total lipids (TLs), total phospholipids (PLs), triacylglycerols (TAGs), cholesterol (Chol), cholesterol esters (Chol esters) and wax esters was analyzed by TLC, the phosphatidylserine (PS), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylcholine (PC), lysophosphatidylcholine (LPC) and sphingomyelin (SM) were determined by HPLC, and fatty acids of total lipids using GC. The lipid sac is a system of cavities filled with lipids, and it is not directly connected to organs of the digestive system. The wall's inner layer is a multinuclear symplast that has a trophic function. The results provide additional knowledge on the role of lipids in the biochemical and physiological adaptation of fish to specific environments and clarify the relationship between fatty acids and the food specialization of postlarvae. Analysis of the fatty acid (FA) profile of TLs in the muscles and lipid sac of daubed shanny pelagic postlarvae showed it to be tissue- and organ-specific, and tightly associated with seasonal variations of environmental factors (temperature conditions and trophic resources).

Improved protein structure prediction using potentials from deep learning.

Protein structure prediction can be used to determine the three-dimensional shape of a protein from its amino acid sequence1. This problem is of fundamental importance as the structure of a protein largely determines its function2; however, protein structures can be difficult to determine experimentally. Considerable progress has recently been made by leveraging genetic information. It is possible to infer which amino acid residues are in contact by analysing covariation in homologous sequences, which aids in the prediction of protein structures3. Here we show that we can train a neural network to make accurate predictions of the distances between pairs of residues, which convey more information about the structure than contact predictions. Using this information, we construct a potential of mean force4 that can accurately describe the shape of a protein. We find that the resulting potential can be optimized by a simple gradient descent algorithm to generate structures without complex sampling procedures. The resulting system, named AlphaFold, achieves high accuracy, even for sequences with fewer homologous sequences. In the recent Critical Assessment of Protein Structure Prediction5 (CASP13)-a blind assessment of the state of the field-AlphaFold created high-accuracy structures (with template modelling (TM) scores6 of 0.7 or higher) for 24 out of 43 free modelling domains, whereas the next best method, which used sampling and contact information, achieved such accuracy for only 14 out of 43 domains. AlphaFold represents a considerable advance in protein-structure prediction. We expect this increased accuracy to enable insights into the function and malfunction of proteins, especially in cases for which no structures for homologous proteins have been experimentally determined7.

Protein structure prediction using multiple deep neural networks in the 13th Critical Assessment of Protein Structure Prediction (CASP13).

We describe AlphaFold, the protein structure prediction system that was entered by the group A7D in CASP13. Submissions were made by three free-modeling (FM) methods which combine the predictions of three neural networks. All three systems were guided by predictions of distances between pairs of residues produced by a neural network. Two systems assembled fragments produced by a generative neural network, one using scores from a network trained to regress GDT_TS. The third system shows that simple gradient descent on a properly constructed potential is able to perform on par with more expensive traditional search techniques and without requiring domain segmentation. In the CASP13 FM assessors' ranking by summed z-scores, this system scored highest with 68.3 vs 48.2 for the next closest group (an average GDT_TS of 61.4). The system produced high-accuracy structures (with GDT_TS scores of 70 or higher) for 11 out of 43 FM domains. Despite not explicitly using template information, the results in the template category were comparable to the best performing template-based methods.

Remote sensing of zooplankton swarms.

Zooplankton provide the key link between primary production and higher levels of the marine food web and they play an important role in mediating carbon sequestration in the ocean. All commercially harvested fish species depend on zooplankton populations. However, spatio-temporal distributions of zooplankton are notoriously difficult to quantify from ships. We know that zooplankton can form large aggregations that visibly change the color of the sea, but the scale and mechanisms producing these features are poorly known. Here we show that large surface patches (>1000 km2) of the red colored copepod Calanus finmarchicus can be identified from satellite observations of ocean color. Such observations provide the most comprehensive view of the distribution of a zooplankton species to date, and alter our understanding of the behavior of this key zooplankton species. Moreover, our findings suggest that high concentrations of astaxanthin-rich zooplankton can degrade the performance of standard blue-green reflectance ratio algorithms in operational use for retrieving chlorophyll concentrations from ocean color remote sensing.

Wireless Communication Technologies for Safe Cooperative Cyber Physical Systems.

Cooperative Cyber-Physical Systems (Co-CPSs) can be enabled using wireless communication technologies, which in principle should address reliability and safety challenges. Safety for Co-CPS enabled by wireless communication technologies is a crucial aspect and requires new dedicated design approaches. In this paper, we provide an overview of five Co-CPS use cases, as introduced in our SafeCOP EU project, and analyze their safety design requirements. Next, we provide a comprehensive analysis of the main existing wireless communication technologies giving details about the protocols developed within particular standardization bodies. We also investigate to what extent they address the non-functional requirements in terms of safety, security and real time, in the different application domains of each use case. Finally, we discuss general recommendations about the use of different wireless communication technologies showing their potentials in the selected real-world use cases. The discussion is provided under consideration in the 5G standardization process within 3GPP, whose current efforts are inline to current gaps in wireless communications protocols for Co-CPSs including many future use cases.

Seasonal ecology in ice-covered Arctic seas - Considerations for spill response decision making.

Due to retreating sea ice and predictions of undiscovered oil and gas resources, increased activity in Arctic shelf sea areas associated with shipping and oil and gas exploration is expected. Such activities may accidentally lead to oil spills in partly ice-covered ocean areas, which raises issues related to oil spill response. Net Environmental Benefit Analysis (NEBA) is the process that the response community uses to identify which combination of response strategies minimises the impact to environment and people. The vulnerability of Valued Ecosystem Components (VEC's) to oil pollution depends on their sensitivity to oil and the likelihood that they will be exposed to oil. As such, NEBA requires a good ecological knowledge base on biodiversity, species' distributions in time and space, and timing of ecological events. Biological resources found at interfaces (e.g., air/water, ice/water or water/coastline) are in general vulnerable because that is where oil can accumulate. Here, we summarize recent information about the seasonal, physical and ecological processes in Arctic waters and evaluate the importance these processes when considering in oil spill response decision making through NEBA. In spring-time, many boreal species conduct a lateral migration northwards in response to sea ice retraction and increased production associated with the spring bloom. However, many Arctic species, including fish, seabirds and marine mammals, are present in upper water layers in the Arctic throughout the year, and recent research has demonstrated that bioactivity during the Arctic winter is higher than previously assumed. Information on the seasonal presence/absence of less resilient VEC's such as marine mammals and sea birds in combination with the presence/absence of sea ice seems to be especially crucial to consider in a NEBA. In addition, quantification of the potential impact of different, realistic spill sizes on the energy cascade following the spring bloom at the ice-edge would provide important information for assessing ecosystem effects.

How including ecological realism impacts the assessment of the environmental effect of oil spills at the population level: The application of matrix models for Arctic Calanus species.

For oil spill responses, assessment of the potential environmental exposure and impacts of a spill is crucial. Due to a lack of chronic toxicity data, acute data is used together with precautionary assumptions. The effect on the Arctic keystone (copepod) species Calanus hyperboreus and Calanus glacialis populations is compared using two approaches: a precautionary approach where all exposed individuals die above a defined threshold concentration and a refined (full-dose-response) approach. For this purpose a matrix population model parameterised with data from the literature is used. Population effects of continuous exposures with varying durations were modelled on a range of concentrations. Just above the chronic No Observed Effect Concentration (which is field relevant) the estimated population recovery duration of the precautionary approach was more than 300 times that of the refined approach. With increasing exposure concentration and duration, the effect in the refined approach converges to the maximum effect assumed in the precautionary approach.

Vector-based navigation using grid-like representations in artificial agents.

Deep neural networks have achieved impressive successes in fields ranging from object recognition to complex games such as Go1,2. Navigation, however, remains a substantial challenge for artificial agents, with deep neural networks trained by reinforcement learning3-5 failing to rival the proficiency of mammalian spatial behaviour, which is underpinned by grid cells in the entorhinal cortex 6 . Grid cells are thought to provide a multi-scale periodic representation that functions as a metric for coding space7,8 and is critical for integrating self-motion (path integration)6,7,9 and planning direct trajectories to goals (vector-based navigation)7,10,11. Here we set out to leverage the computational functions of grid cells to develop a deep reinforcement learning agent with mammal-like navigational abilities. We first trained a recurrent network to perform path integration, leading to the emergence of representations resembling grid cells, as well as other entorhinal cell types 12 . We then showed that this representation provided an effective basis for an agent to locate goals in challenging, unfamiliar, and changeable environments-optimizing the primary objective of navigation through deep reinforcement learning. The performance of agents endowed with grid-like representations surpassed that of an expert human and comparison agents, with the metric quantities necessary for vector-based navigation derived from grid-like units within the network. Furthermore, grid-like representations enabled agents to conduct shortcut behaviours reminiscent of those performed by mammals. Our findings show that emergent grid-like representations furnish agents with a Euclidean spatial metric and associated vector operations, providing a foundation for proficient navigation. As such, our results support neuroscientific theories that see grid cells as critical for vector-based navigation7,10,11, demonstrating that the latter can be combined with path-based strategies to support navigation in challenging environments.

Unexpected Levels of Biological Activity during the Polar Night Offer New Perspectives on a Warming Arctic.

The current understanding of Arctic ecosystems is deeply rooted in the classical view of a bottom-up controlled system with strong physical forcing and seasonality in primary-production regimes. Consequently, the Arctic polar night is commonly disregarded as a time of year when biological activities are reduced to a minimum due to a reduced food supply. Here, based upon a multidisciplinary ecosystem-scale study from the polar night at 79°N, we present an entirely different view. Instead of an ecosystem that has entered a resting state, we document a system with high activity levels and biological interactions across most trophic levels. In some habitats, biological diversity and presence of juvenile stages were elevated in winter months compared to the more productive and sunlit periods. Ultimately, our results suggest a different perspective regarding ecosystem function that will be of importance for future environmental management and decision making, especially at a time when Arctic regions are experiencing accelerated environmental change [1].

Is Ambient Light during the High Arctic Polar Night Sufficient to Act as a Visual Cue for Zooplankton?

The light regime is an ecologically important factor in pelagic habitats, influencing a range of biological processes. However, the availability and importance of light to these processes in high Arctic zooplankton communities during periods of 'complete' darkness (polar night) are poorly studied. Here we characterized the ambient light regime throughout the diel cycle during the high Arctic polar night, and ask whether visual systems of Arctic zooplankton can detect the low levels of irradiance available at this time. To this end, light measurements with a purpose-built irradiance sensor and coupled all-sky digital photographs were used to characterize diel skylight irradiance patterns over 24 hours at 79°N in January 2014 and 2015. Subsequent skylight spectral irradiance and in-water optical property measurements were used to model the underwater light field as a function of depth, which was then weighted by the electrophysiologically determined visual spectral sensitivity of a dominant high Arctic zooplankter, Thysanoessa inermis. Irradiance in air ranged between 1-1.5 x 10-5 µmol photons m-2 s-1 (400-700 nm) in clear weather conditions at noon and with the moon below the horizon, hence values reflect only solar illumination. Radiative transfer modelling generated underwater light fields with peak transmission at blue-green wavelengths, with a 465 nm transmission maximum in shallow water shifting to 485 nm with depth. To the eye of a zooplankter, light from the surface to 75 m exhibits a maximum at 485 nm, with longer wavelengths (>600 nm) being of little visual significance. Our data are the first quantitative characterisation, including absolute intensities, spectral composition and photoperiod of biologically relevant solar ambient light in the high Arctic during the polar night, and indicate that some species of Arctic zooplankton are able to detect and utilize ambient light down to 20-30m depth during the Arctic polar night.

Human-level control through deep reinforcement learning.

The theory of reinforcement learning provides a normative account, deeply rooted in psychological and neuroscientific perspectives on animal behaviour, of how agents may optimize their control of an environment. To use reinforcement learning successfully in situations approaching real-world complexity, however, agents are confronted with a difficult task: they must derive efficient representations of the environment from high-dimensional sensory inputs, and use these to generalize past experience to new situations. Remarkably, humans and other animals seem to solve this problem through a harmonious combination of reinforcement learning and hierarchical sensory processing systems, the former evidenced by a wealth of neural data revealing notable parallels between the phasic signals emitted by dopaminergic neurons and temporal difference reinforcement learning algorithms. While reinforcement learning agents have achieved some successes in a variety of domains, their applicability has previously been limited to domains in which useful features can be handcrafted, or to domains with fully observed, low-dimensional state spaces. Here we use recent advances in training deep neural networks to develop a novel artificial agent, termed a deep Q-network, that can learn successful policies directly from high-dimensional sensory inputs using end-to-end reinforcement learning. We tested this agent on the challenging domain of classic Atari 2600 games. We demonstrate that the deep Q-network agent, receiving only the pixels and the game score as inputs, was able to surpass the performance of all previous algorithms and achieve a level comparable to that of a professional human games tester across a set of 49 games, using the same algorithm, network architecture and hyperparameters. This work bridges the divide between high-dimensional sensory inputs and actions, resulting in the first artificial agent that is capable of learning to excel at a diverse array of challenging tasks.