Skolkovo Institute of Science and Technology

Abstract Cancer is driven by genetic change, and the advent of massively parallel sequencing has enabled systematic documentation of this variation at the whole-genome scale 1–3 . Here we report the integrative analysis of 2,658 whole-cancer genomes and their matching normal tissues across 38 tumour types from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA). We describe the generation of the PCAWG resource, facilitated by international data sharing using compute clouds. On average, cancer genomes contained 4–5 driver mutations when combining coding and non-coding genomic elements; however, in around 5% of cases no drivers were identified, suggesting that cancer driver discovery is not yet complete. Chromothripsis, in which many clustered structural variants arise in a single catastrophic event, is frequently an early event in tumour evolution; in acral melanoma, for example, these events precede most somatic point mutations and affect several cancer-associated genes simultaneously. Cancers with abnormal telomere maintenance often originate from tissues with low replicative activity and show several mechanisms of preventing telomere attrition to critical levels. Common and rare germline variants affect patterns of somatic mutation, including point mutations, structural variants and somatic retrotransposition. A collection of papers from the PCAWG Consortium describes non-coding mutations that drive cancer beyond those in the TERT promoter 4 ; identifies new signatures of mutational processes that cause base substitutions, small insertions and deletions and structural variation 5,6 ; analyses timings and patterns of tumour evolution 7 ; describes the diverse transcriptional consequences of somatic mutation on splicing, expression levels, fusion genes and promoter activity 8,9 ; and evaluates a range of more-specialized features of cancer genomes 8,10–18 .

At the atomic-cluster scale, pure boron is markedly similar to carbon, forming simple planar molecules and cage-like fullerenes. Theoretical studies predict that two-dimensional (2D) boron sheets will adopt an atomic configuration similar to that of boron atomic clusters. We synthesized atomically thin, crystalline 2D boron sheets (i.e., borophene) on silver surfaces under ultrahigh-vacuum conditions. Atomic-scale characterization, supported by theoretical calculations, revealed structures reminiscent of fused boron clusters with multiple scales of anisotropic, out-of-plane buckling. Unlike bulk boron allotropes, borophene shows metallic characteristics that are consistent with predictions of a highly anisotropic, 2D metal.

Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of "deep" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation. Overall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets.

INTRODUCTION Almost all archaea and about half of bacteria possess clustered regularly interspaced short palindromic repeat (CRISPR)–CRISPR-associated genes (Cas) adaptive immune systems, which protect microbes against viruses and other foreign DNA. All functionally characterized CRISPR systems have been reported to target DNA, with some multicomponent type III systems also targeting RNA. The putative class 2 type VI system, which has not been functionally characterized, encompasses the single-effector protein C2c2, which contains two Higher Eukaryotes and Prokaryotes Nucleotide-binding (HEPN) domains commonly associated with ribonucleases (RNases), suggesting RNA-guided RNA-targeting function. RATIONALE Existing studies have only established a role for RNA interference, in addition to DNA interference, in the multicomponent type III-A and III-B systems. We investigated the possibility of C2c2-mediated RNA inference by heterologously expressing C2c2 locus from Leptotrichia shahii (LshC2c2) in the model system Escherichia coli. The ability of LshC2c2 to protect against MS2 single-stranded RNA (ssRNA) phage infection was assessed by using every possible spacer sequence against the phage genome. We next developed protocols to reconstitute purified recombinant LshC2c2 protein and test its biochemical activity when incubated with its mature CRISPR RNA (crRNA) and target ssRNA. We systematically evaluated the parameters necessary for cleavage. Last, to demonstrate the potential utility of the LshC2c2 complex for RNA targeting in living bacterial cells, we guided it to knockdown red fluorescent protein (RFP) mRNA in vivo. RESULTS This work demonstrates the RNA-guided RNase activity of the putative type VI CRISPR-effector LshC2c2. Heterologously expressed C2c2 can protect E. coli from MS2 phage, and by screening against the MS2 genome, we identified a H (non-G) protospacer flanking site (PFS) following the RNA target site, which was confirmed by targeting a complementary sequence in the β-lactamase transcript followed by a degenerate nucleotide sequence. Using purified LshC2c2 protein, we demonstrate that C2c2 and crRNA are sufficient in vitro to achieve RNA-guided, PFS-dependent RNA cleavage. This cleavage preferentially occurs at uracil residues in ssRNA regions and depends on conserved catalytic residues in the two HEPN domains. Mutation of these residues yields a catalytically inactive RNA-binding protein. The secondary structure of the crRNA direct repeat (DR) stem is required for LshC2c2 activity, and mutations in the 3′ region of the DR eliminate cleavage activity. Targeting is also sensitive to multiple or consecutive mismatches in the spacer:protospacer duplex. C2c2 targeting of RFP mRNA in vivo results in reduced fluorescence. The knockdown of the RFP mRNA by C2c2 slowed E. coli growth, and in agreement with this finding, in vitro cleavage of the target RNA results in “collateral,” nonspecific cleavage of other RNAs present in the reaction mix. CONCLUSION LshC2c2 is a RNA-guided RNase which requires the activity of its two HEPN domains, suggesting previously unidentified mechanisms of RNA targeting and degradation by CRISPR systems. Promiscuous RNase activity of C2c2 after activation by the target slows bacterial growth and suggests that C2c2 could protect bacteria from virus spread via programmed cell death and dormancy induction. A single-effector RNA targeting system has the potential to serve as a general chassis for molecular tools for visualizing, degrading, or binding RNA in a programmable, multiplexed fashion. C2c2 is an RNA-guided RNase that provides protection against RNA phage. CRISPR-C2c2 from L. shahii can be reconstituted in E. coli to mediate RNA-guided interference of the RNA phage MS2. Biochemical characterization of C2c2 reveals crRNA-guided RNA cleavage facilitated by the two HEPN nuclease domains. Binding of the target RNA by C2c2-crRNA also activates a nonspecific RNase activity, which may lead to promiscuous cleavage of RNAs without complementarity to the crRNA guide sequence.

CellProfiler has enabled the scientific research community to create flexible, modular image analysis pipelines since its release in 2005. Here, we describe CellProfiler 3.0, a new version of the software supporting both whole-volume and plane-wise analysis of three-dimensional (3D) image stacks, increasingly common in biomedical research. CellProfiler's infrastructure is greatly improved, and we provide a protocol for cloud-based, large-scale image processing. New plugins enable running pretrained deep learning models on images. Designed by and for biologists, CellProfiler equips researchers with powerful computational tools via a well-documented user interface, empowering biologists in all fields to create quantitative, reproducible image analysis workflows.

OBJECTIVE: Most current electroencephalography (EEG)-based brain-computer interfaces (BCIs) are based on machine learning algorithms. There is a large diversity of classifier types that are used in this field, as described in our 2007 review paper. Now, approximately ten years after this review publication, many new algorithms have been developed and tested to classify EEG signals in BCIs. The time is therefore ripe for an updated review of EEG classification algorithms for BCIs. APPROACH: We surveyed the BCI and machine learning literature from 2007 to 2017 to identify the new classification approaches that have been investigated to design BCIs. We synthesize these studies in order to present such algorithms, to report how they were used for BCIs, what were the outcomes, and to identify their pros and cons. MAIN RESULTS: We found that the recently designed classification algorithms for EEG-based BCIs can be divided into four main categories: adaptive classifiers, matrix and tensor classifiers, transfer learning and deep learning, plus a few other miscellaneous classifiers. Among these, adaptive classifiers were demonstrated to be generally superior to static ones, even with unsupervised adaptation. Transfer learning can also prove useful although the benefits of transfer learning remain unpredictable. Riemannian geometry-based methods have reached state-of-the-art performances on multiple BCI problems and deserve to be explored more thoroughly, along with tensor-based methods. Shrinkage linear discriminant analysis and random forests also appear particularly useful for small training samples settings. On the other hand, deep learning methods have not yet shown convincing improvement over state-of-the-art BCI methods. SIGNIFICANCE: This paper provides a comprehensive overview of the modern classification algorithms used in EEG-based BCIs, presents the principles of these methods and guidelines on when and how to use them. It also identifies a number of challenges to further advance EEG classification in BCI.

Deep convolutional networks have become a popular tool for image generation and restoration. Generally, their excellent performance is imputed to their ability to learn realistic image priors from a large number of example images. In this paper, we show that, on the contrary, the structure of a generator network is sufficient to capture a great deal of low-level image statistics prior to any learning. In order to do so, we show that a randomly-initialized neural network can be used as a handcrafted prior with excellent results in standard inverse problems such as denoising, superresolution, and inpainting. Furthermore, the same prior can be used to invert deep neural representations to diagnose them, and to restore images based on flash-no flash input pairs. Apart from its diverse applications, our approach highlights the inductive bias captured by standard generator network architectures. It also bridges the gap between two very popular families of image restoration methods: learning-based methods using deep convolutional networks and learning-free methods based on handcrafted image priors such as self-similarity.

Abstract Improving the long-term stability of perovskite solar cells is critical to the deployment of this technology. Despite the great emphasis laid on stability-related investigations, publications lack consistency in experimental procedures and parameters reported. It is therefore challenging to reproduce and compare results and thereby develop a deep understanding of degradation mechanisms. Here, we report a consensus between researchers in the field on procedures for testing perovskite solar cell stability, which are based on the International Summit on Organic Photovoltaic Stability (ISOS) protocols. We propose additional procedures to account for properties specific to PSCs such as ion redistribution under electric fields, reversible degradation and to distinguish ambient-induced degradation from other stress factors. These protocols are not intended as a replacement of the existing qualification standards, but rather they aim to unify the stability assessment and to understand failure modes. Finally, we identify key procedural information which we suggest reporting in publications to improve reproducibility and enable large data set analysis.

Density functional theory offers a very accurate way of computing materials properties from first principles. However, it is too expensive for modeling large-scale molecular systems whose properties are, in contrast, computed using interatomic potentials. The present paper considers, from a mathematical point of view, the problem of constructing interatomic potentials that approximate a given quantum-mechanical interaction model. In particular, a new class of systematically improvable potentials is proposed, analyzed, and tested on an existing quantum-mechanical database.

Perovskite oxides are attractive candidates as catalysts for the electrolysis of water in alkaline energy storage and conversion systems. However, the rational design of active catalysts has been hampered by the lack of understanding of the mechanism of water electrolysis on perovskite surfaces. Key parameters that have been overlooked include the role of oxygen vacancies, B-O bond covalency, and redox activity of lattice oxygen species. Here we present a series of cobaltite perovskites where the covalency of the Co-O bond and the concentration of oxygen vacancies are controlled through Sr(2+) substitution into La(1-x)Sr(x)CoO(3-δ) . We attempt to rationalize the high activities of La(1-x)Sr(x)CoO(3-δ) through the electronic structure and participation of lattice oxygen in the mechanism of water electrolysis as revealed through ab initio modelling. Using this approach, we report a material, SrCoO2.7, with a high, room temperature-specific activity and mass activity towards alkaline water electrolysis.

We have devised and implemented a novel computational strategy for de novo design of molecules with desired properties termed ReLeaSE (Reinforcement Learning for Structural Evolution). On the basis of deep and reinforcement learning (RL) approaches, ReLeaSE integrates two deep neural networks-generative and predictive-that are trained separately but are used jointly to generate novel targeted chemical libraries. ReLeaSE uses simple representation of molecules by their simplified molecular-input line-entry system (SMILES) strings only. Generative models are trained with a stack-augmented memory network to produce chemically feasible SMILES strings, and predictive models are derived to forecast the desired properties of the de novo-generated compounds. In the first phase of the method, generative and predictive models are trained separately with a supervised learning algorithm. In the second phase, both models are trained jointly with the RL approach to bias the generation of new chemical structures toward those with the desired physical and/or biological properties. In the proof-of-concept study, we have used the ReLeaSE method to design chemical libraries with a bias toward structural complexity or toward compounds with maximal, minimal, or specific range of physical properties, such as melting point or hydrophobicity, or toward compounds with inhibitory activity against Janus protein kinase 2. The approach proposed herein can find a general use for generating targeted chemical libraries of novel compounds optimized for either a single desired property or multiple properties.

We present a major update of the HOCOMOCO collection that consists of patterns describing DNA binding specificities for human and mouse transcription factors. In this release, we profited from a nearly doubled volume of published in vivo experiments on transcription factor (TF) binding to expand the repertoire of binding models, replace low-quality models previously based on in vitro data only and cover more than a hundred TFs with previously unknown binding specificities. This was achieved by systematic motif discovery from more than five thousand ChIP-Seq experiments uniformly processed within the BioUML framework with several ChIP-Seq peak calling tools and aggregated in the GTRD database. HOCOMOCO v11 contains binding models for 453 mouse and 680 human transcription factors and includes 1302 mononucleotide and 576 dinucleotide position weight matrices, which describe primary binding preferences of each transcription factor and reliable alternative binding specificities. An interactive interface and bulk downloads are available on the web: http://hocomoco.autosome.ru and http://www.cbrc.kaust.edu.sa/hocomoco11. In this release, we complement HOCOMOCO by MoLoTool (Motif Location Toolbox, http://molotool.autosome.ru) that applies HOCOMOCO models for visualization of binding sites in short DNA sequences.

Abstract A key mutational process in cancer is structural variation, in which rearrangements delete, amplify or reorder genomic segments that range in size from kilobases to whole chromosomes 1–7 . Here we develop methods to group, classify and describe somatic structural variants, using data from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA), which aggregated whole-genome sequencing data from 2,658 cancers across 38 tumour types 8 . Sixteen signatures of structural variation emerged. Deletions have a multimodal size distribution, assort unevenly across tumour types and patients, are enriched in late-replicating regions and correlate with inversions. Tandem duplications also have a multimodal size distribution, but are enriched in early-replicating regions—as are unbalanced translocations. Replication-based mechanisms of rearrangement generate varied chromosomal structures with low-level copy-number gains and frequent inverted rearrangements. One prominent structure consists of 2–7 templates copied from distinct regions of the genome strung together within one locus. Such cycles of templated insertions correlate with tandem duplications, and—in liver cancer—frequently activate the telomerase gene TERT . A wide variety of rearrangement processes are active in cancer, which generate complex configurations of the genome upon which selection can act.

Modern image inpainting systems, despite the significant progress, often struggle with large missing areas, complex geometric structures, and high-resolution images. We find that one of the main reasons for that is the lack of an effective receptive field in both the inpainting network and the loss function. To alleviate this issue, we propose a new method called large mask inpainting (LaMa). LaMa is based on i) a new inpainting network architecture that uses fast Fourier convolutions (FFCs), which have the image-wide receptive field; ii) a high receptive field perceptual loss; iii) large training masks, which unlocks the potential of the first two components. Our inpainting network improves the state-of-the-art across a range of datasets and achieves excellent performance even in challenging scenarios, e.g. completion of periodic structures. Our model generalizes surprisingly well to resolutions that are higher than those seen at train time, and achieves this at lower parameter&time costs than the competitive baselines. The code is available at https://github.com/saic-mdal/lama.

Abstract Hydrogen-rich superhydrides are believed to be very promising high-Tc superconductors. Recent experiments discovered superhydrides at very high pressures, e.g. FeH5 at 130 GPa and LaH10 at 170 GPa. With the motivation of discovering new hydrogen-rich high-Tc superconductors at lowest possible pressure, here we report the prediction and experimental synthesis of cerium superhydride CeH9 at 80–100 GPa in the laser-heated diamond anvil cell coupled with synchrotron X-ray diffraction. Ab initio calculations were carried out to evaluate the detailed chemistry of the Ce-H system and to understand the structure, stability and superconductivity of CeH9. CeH9 crystallizes in a P63/mmc clathrate structure with a very dense 3-dimensional atomic hydrogen sublattice at 100 GPa. These findings shed a significant light on the search for superhydrides in close similarity with atomic hydrogen within a feasible pressure range. Discovery of superhydride CeH9 provides a practical platform to further investigate and understand conventional superconductivity in hydrogen rich superhydrides.

Machine learning of the quantitative relationship between local environment descriptors and the potential energy surface of a system of atoms has emerged as a new frontier in the development of interatomic potentials (IAPs). Here, we present a comprehensive evaluation of machine learning IAPs (ML-IAPs) based on four local environment descriptors-atom-centered symmetry functions (ACSF), smooth overlap of atomic positions (SOAP), the spectral neighbor analysis potential (SNAP) bispectrum components, and moment tensors-using a diverse data set generated using high-throughput density functional theory (DFT) calculations. The data set comprising bcc (Li, Mo) and fcc (Cu, Ni) metals and diamond group IV semiconductors (Si, Ge) is chosen to span a range of crystal structures and bonding. All descriptors studied show excellent performance in predicting energies and forces far surpassing that of classical IAPs, as well as predicting properties such as elastic constants and phonon dispersion curves. We observe a general trade-off between accuracy and the degrees of freedom of each model and, consequently, computational cost. We will discuss these trade-offs in the context of model selection for molecular dynamics and other applications.

The recent work of Gatys et al., who characterized the style of an image by the statistics of convolutional neural network filters, ignited a renewed interest in the texture generation and image stylization problems. While their image generation technique uses a slow optimization process, recently several authors have proposed to learn generator neural networks that can produce similar outputs in one quick forward pass. While generator networks are promising, they are still inferior in visual quality and diversity compared to generation-by-optimization. In this work, we advance them in two significant ways. First, we introduce an instance normalization module to replace batch normalization with significant improvements to the quality of image stylization. Second, we improve diversity by introducing a new learning formulation that encourages generators to sample unbiasedly from the Julesz texture ensemble, which is the equivalence class of all images characterized by certain filter responses. Together, these two improvements take feed forward texture synthesis and image stylization much closer to the quality of generation-via-optimization, while retaining the speed advantage.

Lithium-ion (Li-ion) batteries that rely on cationic redox reactions are the primary energy source for portable electronics. One pathway toward greater energy density is through the use of Li-rich layered oxides. The capacity of this class of materials (>270 milliampere hours per gram) has been shown to be nested in anionic redox reactions, which are thought to form peroxo-like species. However, the oxygen-oxygen (O-O) bonding pattern has not been observed in previous studies, nor has there been a satisfactory explanation for the irreversible changes that occur during first delithiation. By using Li2IrO3 as a model compound, we visualize the O-O dimers via transmission electron microscopy and neutron diffraction. Our findings establish the fundamental relation between the anionic redox process and the evolution of the O-O bonding in layered oxides.

We show that the time-dependent variational principle provides a unifying framework for time-evolution methods and optimization methods in the context of matrix product states. In particular, we introduce a new integration scheme for studying time evolution, which can cope with arbitrary Hamiltonians, including those with long-range interactions. Rather than a Suzuki-Trotter splitting of the Hamiltonian, which is the idea behind the adaptive time-dependent density matrix renormalization group method or time-evolving block decimation, our method is based on splitting the projector onto the matrix product state tangent space as it appears in the Dirac-Frenkel time-dependent variational principle. We discuss how the resulting algorithm resembles the density matrix renormalization group (DMRG) algorithm for finding ground states so closely that it can be implemented by changing just a few lines of code and it inherits the same stability and efficiency. In particular, our method is compatible with any Hamiltonian for which ground-state DMRG can be implemented efficiently. In fact, DMRG is obtained as a special case of our scheme for imaginary time evolution with infinite time step.

Covering: up to June 2020Ribosomally-synthesized and post-translationally modified peptides (RiPPs) are a large group of natural products. A community-driven review in 2013 described the emerging commonalities in the biosynthesis of RiPPs and the opportunities they offered for bioengineering and genome mining. Since then, the field has seen tremendous advances in understanding of the mechanisms by which nature assembles these compounds, in engineering their biosynthetic machinery for a wide range of applications, and in the discovery of entirely new RiPP families using bioinformatic tools developed specifically for this compound class. The First International Conference on RiPPs was held in 2019, and the meeting participants assembled the current review describing new developments since 2013. The review discusses the new classes of RiPPs that have been discovered, the advances in our understanding of the installation of both primary and secondary post-translational modifications, and the mechanisms by which the enzymes recognize the leader peptides in their substrates. In addition, genome mining tools used for RiPP discovery are discussed as well as various strategies for RiPP engineering. An outlook section presents directions for future research.