Explore the vast range of titles available.

Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socioeconomic needs of many sectors reliant on weather-dependent decision-making 1 , 2 . State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-based wind estimates, and struggle to capture important non-linear events such as convective initiations 3 , 4 . Recently introduced deep learning methods use radar to directly predict future rain rates, free of physical constraints 5 , 6 . While they accurately predict low-intensity rainfall, their operational utility is limited because their lack of constraints produces blurry nowcasts at longer lead times, yielding poor performance on rarer medium-to-heavy rain events. Here we present a deep generative model for the probabilistic nowcasting of precipitation from radar that addresses these challenges. Using statistical, economic and cognitive measures, we show that our method provides improved forecast quality, forecast consistency and forecast value. Our model produces realistic and spatiotemporally consistent predictions over regions up to 1,536 km × 1,280 km and with lead times from 5–90 min ahead. Using a systematic evaluation by more than 50 expert meteorologists, we show that our generative model ranked first for its accuracy and usefulness in 89% of cases against two competitive methods. When verified quantitatively, these nowcasts are skillful without resorting to blurring. We show that generative nowcasting can provide probabilistic predictions that improve forecast value and support operational utility, and at resolutions and lead times where alternative methods struggle. A deep generative model using radar observations is used to create skilful precipitation predictions that are accurate and support real-world utility.

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 structure 1 . Here we markedly expand the structural coverage of the proteome by applying the state-of-the-art machine learning method, AlphaFold 2 , 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. AlphaFold is used to predict the structures of almost all of the proteins in the human proteome—the availability of high-confidence predicted structures could enable new avenues of investigation from a structural perspective.

Cytoplasmic viscosity is a crucial parameter in determining rates of diffusion-limited reactions. Changes in viscosity are associated with several diseases, whilst nuclear viscosity determines gene integrity, regulation and expression. Yet how drugs including DNA-damaging agents affect viscosity is unknown. We demonstrate the use of a platinum complex, Pt[L]Cl, that localizes efficiently mostly in the nucleus as a probe for nuclear viscosity. The phosphorescence lifetime of Pt[L]Cl is sensitive to viscosity and provides an excellent tool to investigate the impact of DNA damage. We show using Fluorescence Lifetime Imaging (FLIM) that the lifetime of both green and red fluorescent proteins (FP) are also sensitive to changes in cellular viscosity and refractive index. However, Pt[L]Cl proved to be a more sensitive viscosity probe, by virtue of microsecond phosphorescence lifetime versus nanosecond fluorescence lifetime of FP, hence greater sensitivity to bimolecular reactions. DNA damage was inflicted by either a two-photon excitation, one-photon excitation microbeam and X-rays. DNA damage of live cells causes significant increase in the lifetime of either Pt[L]Cl (HeLa cells, 12.5–14.1 µs) or intracellularly expressed mCherry (HEK293 cells, 1.54–1.67 ns), but a decrease in fluorescence lifetime of GFP from 2.65 to 2.29 ns (in V15B cells). These values represent a viscosity change from 8.59 to 20.56 cP as well as significant changes in the refractive index (RI), according to independent calibration. Interestingly DNA damage localized to a submicron region following a laser microbeam induction showed a whole cell viscosity change, with those in the nucleus being greater than the cytoplasm. We also found evidence of a by-stander effect, whereby adjacent un-irradiated cells also showed nuclear viscosity change. Finally, an increase in viscosity following DNA damage was also observed in bacterial cells with an over-expressed mNeonGreen FP, evidenced by the change in its lifetime from 2.8 to 2.4 ns.

Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort 1 , 2 , 3 – 4 , the structures of around 100,000 unique proteins have been determined 5 , but this represents a small fraction of the billions of known protein sequences 6 , 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 years 9 . Despite recent progress 10 , 11 , 12 , 13 – 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. AlphaFold predicts protein structures with an accuracy competitive with experimental structures in the majority of cases using a novel deep learning architecture.

Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socio-economic needs of many sectors reliant on weather-dependent decision-making. State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-based wind estimates, and struggle to capture important non-linear events such as convective initiations. Recently introduced deep learning methods use radar to directly predict future rain rates, free of physical constraints. While they accurately predict low-intensity rainfall, their operational utility is limited because their lack of constraints produces blurry nowcasts at longer lead times, yielding poor performance on more rare medium-to-heavy rain events. To address these challenges, we present a Deep Generative Model for the probabilistic nowcasting of precipitation from radar. Our model produces realistic and spatio-temporally consistent predictions over regions up to 1536 km x 1280 km and with lead times from 5-90 min ahead. In a systematic evaluation by more than fifty expert forecasters from the Met Office, our generative model ranked first for its accuracy and usefulness in 88% of cases against two competitive methods, demonstrating its decision-making value and ability to provide physical insight to real-world experts. When verified quantitatively, these nowcasts are skillful without resorting to blurring. We show that generative nowcasting can provide probabilistic predictions that improve forecast value and support operational utility, and at resolutions and lead times where alternative methods struggle.

We describe the 2020 edition of the DeepMind Kinetics human action dataset, which replenishes and extends the Kinetics-700 dataset. In this new version, there are at least 700 video clips from different YouTube videos for each of the 700 classes. This paper details the changes introduced for this new release of the dataset and includes a comprehensive set of statistics as well as baseline results using the I3D network.

Cryo-electron microscopy (cryo-EM) has revolutionized experimental protein structure determination. Despite advances in high resolution reconstruction, a majority of cryo-EM experiments provide either a single state of the studied macromolecule, or a relatively small number of its conformations. This reduces the effectiveness of the technique for proteins with flexible regions, which are known to play a key role in protein function. Recent methods for capturing conformational heterogeneity in cryo-EM data model it in volume space, making recovery of continuous atomic structures challenging. Here we present a fully deep-learning-based approach using variational auto-encoders (VAEs) to recover a continuous distribution of atomic protein structures and poses directly from picked particle images and demonstrate its efficacy on realistic simulated data. We hope that methods built on this work will allow incorporation of stronger prior information about protein structure and enable better understanding of non-rigid protein structures.

While the vast majority of well-structured single protein chains can now be predicted to high accuracy due to the recent AlphaFold [1] model, the prediction of multi-chain protein complexes remains a challenge in many cases. In this work, we demonstrate that an AlphaFold model trained specifically for multimeric inputs of known stoichiometry, which we call AlphaFold-Multimer, significantly increases accuracy of predicted multimeric interfaces over input-adapted single-chain AlphaFold while maintaining high intra-chain accuracy. On a benchmark dataset of 17 heterodimer proteins without templates (introduced in [2]) we achieve at least medium accuracy (DockQ [3]≥0.49) on 13 targets and high accuracy (DockQ≥0.8) on 7 targets, compared to 9 targets of at least medium accuracy and 4 of high accuracy for the previous state of the art system (an AlphaFold-based system from [2]). We also predict structures for a large dataset of 4,446 recent protein complexes, from which we score all non-redundant interfaces with low template identity. For heteromeric interfaces we successfully predict the interface (DockQ≥0.23) in 70% of cases, and produce high accuracy predictions (DockQ≥0.8) in 26% of cases, an improvement of +27 and +14 percentage points over the flexible linker modification of AlphaFold [4] respectively. For homomeric interfaces we successfully predict the interface in 72% of cases, and produce high accuracy predictions in 36% of cases, an improvement of +8 and +7 percentage points respectively. Competing Interest Statement The authors have declared no competing interest. Footnotes * This revision updates the results for new models trained with a between chain centre-of-mass loss, re-weighting of the violation losses and removal of the prokaryote specific MSA pairing. These changes significantly reduce the number of structures with clashes and improve the overall accuracy.

The study compared two groups of high school algebra classes at Harlem High School in Harlem, Georgia during the 1992-93 school year. There were 7 instructors, 5 teaching algebra in the traditional textbook method (the control group) and 2 using Learning Logic (the experimental group). The software package was developed for the National Science Center Foundation, Incorporated. Data were collected for all students including pretest/posttest scores, gender, socio-economic background, ethnicity, and enrollment in higher mathematics classes during the 1994-95 school year. Analysis of pretest scores determined that both groups were comparable in prior mathematics knowledge. A significant difference was found in gender and ethnicity on one posttest and in socio-economic background on both posttests. Differences in posttests may indicate gender and ethnic bias. No significant difference was found in delivery method. A significant difference was found in student enrollment in higher mathematics classes by students from the Learning Logic delivery method.

Fifteen women in each of three age groups, 20 to 25, 40 to 45, and 70 to 75 years, walked balance beams under conditions in which both visual and proprioceptive-vestibular (P-V) information varied. P-V conditions included high (3.8 cm beam width), moderate (8.9 cm), and low P-V demands (13.9 cm). The visual conditions included: (1) a normal (Full) visual environment, (2) two reduced visual environments, one in which individuals walked the beams with only a central visual cue (luminescent rod placed beyond the end of the beam) and one in which both central and peripheral visual cues were present (two luminescent rods placed parallel to the beam), and (3) a dark environment. Subjects walked each of the beams three times under each of four visual conditions for a total of 36 trials. Means and standard deviations of total time to walk the beam, percent time-in balance, and number of step-offs were analyzed using a repeated measures MANOVA with appropriate follow up tests. The design was an Age (3) x Visual Condition (4) x P-V Condition (3) factorial with Age as the between subjects factor and the two latter factors as within subject factors. In general, balance performance declined with age, with diminished visual information, and with increased P-V demand. Seventy year olds tended to exhibit a slowing of balance speed but not balance efficiency. Reduced visual information (dark) had a more profound effect on the performance of 40 year olds than 70 year olds. All individuals, regardless of age, were more variable in controlling balance performance when P-V demands were high. Balance control under these conditions varied as a function of the amount of visual information available. With age, increasing P-V demand beyond a minimum level appears to place a demand on the system that is best handled when a full complement of visual cues is available. Vision may be necessary to monitor the use of proprioceptive-vestibular information particularly with a precision (high P-V demand) balance task.