Abstract: How do humans recognize the action “opening a book”? We argue that there are two important cues: modeling temporal shape dynamics and modeling functional relationships between humans and objects. In this paper, we propose to represent videos as space-time region graphs which capture these two important cues. Our graph nodes are defined by the object region proposals from different frames in a long range video. These nodes are connected by two types of relations: (i) similarity relations capturing the long range dependencies between correlated objects and (ii) spatial-temporal relations capturing the interactions between nearby objects. We perform reasoning on this graph representation via Graph Convolutional Networks. We achieve state-of-the-art results on the Charades and Something-Something datasets. Especially for Charades with complex environments, we obtain a huge \(4.4\%\) gain when our model is applied in complex environments.

Abstract: Summary Background Malignant cells of classical Hodgkin's lymphoma are characterised by genetic alterations at the 9p24.1 locus, leading to overexpression of PD-1 ligands and evasion of immune surveillance. In a phase 1b study, nivolumab, a PD-1-blocking antibody, produced a high response in patients with relapsed and refractory classical Hodgkin's lymphoma, with an acceptable safety profile. We aimed to assess the clinical benefit and safety of nivolumab monotherapy in patients with classical Hodgkin's lymphoma after failure of both autologous stem-cell transplantation and brentuximab vedotin. Methods In this ongoing, single-arm phase 2 study, adult patients (aged ≥18 years) with recurrent classical Hodgkin's lymphoma who had failed to respond to autologous stem-cell transplantation and had either relapsed after or failed to respond to brentuximab vedotin, and with an Eastern Cooperative Oncology Group performance status score of 0 or 1, were enrolled from 34 hospitals and academic centres across Europe and North America. Patients were given nivolumab intravenously over 60 min at 3 mg/kg every 2 weeks until progression, death, unacceptable toxicity, or withdrawal from study. The primary endpoint was objective response following a prespecified minimum follow-up period of 6 months, assessed by an independent radiological review committee (IRRC). All patients who received at least one dose of nivolumab were included in the primary and safety analyses. This trial is registered with ClinicalTrials.gov, number NCT02181738. Findings Among 80 treated patients recruited between Aug 26, 2014, and Feb 20, 2015, the median number of previous therapies was four (IQR 4–7). At a median follow-up of 8·9 months (IQR 7·8–9·9), 53 (66·3%, 95% CI 54·8–76·4) of 80 patients achieved an IRRC-assessed objective response. The most common drug-related adverse events (those that occurred in ≥15% of patients) included fatigue (20 [25%] patients), infusion-related reaction (16 [20%]), and rash (13 [16%]). The most common drug-related grade 3 or 4 adverse events were neutropenia (four [5%] patients) and increased lipase concentrations (four [5%]). The most common serious adverse event (any grade) was pyrexia (three [4%] patients). Three patients died during the study; none of these deaths were judged to be treatment related. Interpretation Nivolumab resulted in frequent responses with an acceptable safety profile in patients with classical Hodgkin's lymphoma who progressed after autologous stem-cell transplantation and brentuximab vedotin. Therefore, nivolumab might be a new treatment option for a patient population with a high unmet need. Ongoing follow-up will help to assess the durability of response. Funding Bristol-Myers Squibb.

Abstract: Mitochondria undergo fragmentation in response to electron transport chain (ETC) poisons and mitochondrial DNA–linked disease mutations, yet how these stimuli mechanistically connect to the mitochondrial fission and fusion machinery is poorly understood. We found that the energy-sensing adenosine monophosphate (AMP)–activated protein kinase (AMPK) is genetically required for cells to undergo rapid mitochondrial fragmentation after treatment with ETC inhibitors. Moreover, direct pharmacological activation of AMPK was sufficient to rapidly promote mitochondrial fragmentation even in the absence of mitochondrial stress. A screen for substrates of AMPK identified mitochondrial fission factor (MFF), a mitochondrial outer-membrane receptor for DRP1, the cytoplasmic guanosine triphosphatase that catalyzes mitochondrial fission. Nonphosphorylatable and phosphomimetic alleles of the AMPK sites in MFF revealed that it is a key effector of AMPK-mediated mitochondrial fission.

Abstract: Mitogen-activated protein kinase (MAPK) cascades are key signalling pathways that regulate a wide variety of cellular processes, including proliferation, differentiation, apoptosis and stress responses. The MAPK pathway includes three main kinases, MAPK kinase kinase, MAPK kinase and MAPK, which activate and phosphorylate downstream proteins. The extracellular signal-regulated kinases ERK1 and ERK2 are evolutionarily conserved, ubiquitous serine-threonine kinases that regulate cellular signalling under both normal and pathological conditions. ERK expression is critical for development and their hyperactivation plays a major role in cancer development and progression. The Ras/Raf/MAPK (MEK)/ERK pathway is the most important signalling cascade among all MAPK signal transduction pathways, and plays a crucial role in the survival and development of tumour cells. The present review discusses recent studies on Ras and ERK pathway members. With respect to processes downstream of ERK activation, the role of ERK in tumour proliferation, invasion and metastasis is highlighted, and the role of the ERK/MAPK signalling pathway in tumour extracellular matrix degradation and tumour angiogenesis is emphasised.

Abstract: Human coronavirus (HCoV) infection causes respiratory diseases with mild to severe outcomes. In the last 15 years, we have witnessed the emergence of two zoonotic, highly pathogenic HCoVs: severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV). Replication of HCoV is regulated by a diversity of host factors and induces drastic alterations in cellular structure and physiology. Activation of critical signaling pathways during HCoV infection modulates the induction of antiviral immune response and contributes to the pathogenesis of HCoV. Recent studies have begun to reveal some fundamental aspects of the intricate HCoV-host interaction in mechanistic detail. In this review, we summarize the current knowledge of host factors co-opted and signaling pathways activated during HCoV infection, with an emphasis on HCoV-infection-induced stress response, autophagy, apoptosis, and innate immunity. The cross talk among these pathways, as well as the modulatory strategies utilized by HCoV, is also discussed.

Abstract: The past decade has seen an explosion in the amount of digital information stored in electronic health records (EHRs). While primarily designed for archiving patient information and performing administrative healthcare tasks like billing, many researchers have found secondary use of these records for various clinical informatics applications. Over the same period, the machine learning community has seen widespread advances in the field of deep learning. In this review, we survey the current research on applying deep learning to clinical tasks based on EHR data, where we find a variety of deep learning techniques and frameworks being applied to several types of clinical applications including information extraction, representation learning, outcome prediction, phenotyping, and deidentification. We identify several limitations of current research involving topics such as model interpretability, data heterogeneity, and lack of universal benchmarks. We conclude by summarizing the state of the field and identifying avenues of future deep EHR research.

Abstract: We present a method for extracting depth information from a rectified image pair. We train a convolutional neural network to predict how well two image patches match and use it to compute the stereo matching cost. The cost is refined by cross-based cost aggregation and semiglobal matching, followed by a left-right consistency check to eliminate errors in the occluded regions. Our stereo method achieves an error rate of 2.61% on the KITTI stereo dataset and is currently (August 2014) the top performing method on this dataset.

Abstract: The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fourier space, allowing for an expressive and efficient architecture. We perform experiments on Burgers' equation, Darcy flow, and Navier-Stokes equation. The Fourier neural operator is the first ML-based method to successfully model turbulent flows with zero-shot super-resolution. It is up to three orders of magnitude faster compared to traditional PDE solvers. Additionally, it achieves superior accuracy compared to previous learning-based solvers under fixed resolution.

Abstract: We address the problem of activity detection in continuous, untrimmed video streams. This is a difficult task that requires extracting meaningful spatio-temporal features to capture activities, accurately localizing the start and end times of each activity. We introduce a new model, Region Convolutional 3D Network (R-C3D), which encodes the video streams using a three-dimensional fully convolutional network, then generates candidate temporal regions containing activities, and finally classifies selected regions into specific activities. Computation is saved due to the sharing of convolutional features between the proposal and the classification pipelines. The entire model is trained end-to-end with jointly optimized localization and classification losses. R-C3D is faster than existing methods (569 frames per second on a single Titan X Maxwell GPU) and achieves state-of-the-art results on THUMOS’14. We further demonstrate that our model is a general activity detection framework that does not rely on assumptions about particular dataset properties by evaluating our approach on ActivityNet and Charades. Our code is available at http://ai.bu.edu/r-c3d/

Abstract: Pyroptosis is a lytic type of cell death that is initiated by inflammatory caspases. These caspases are activated within multi-protein inflammasome complexes that assemble in response to pathogens and endogenous danger signals. Pyroptotic cell death has been proposed to proceed via the formation of a plasma membrane pore, but the underlying molecular mechanism has remained unclear. Recently, gasdermin D (GSDMD), a member of the ill-characterized gasdermin protein family, was identified as a caspase substrate and an essential mediator of pyroptosis. GSDMD is thus a candidate for pyroptotic pore formation. Here, we characterize GSDMD function in live cells and in vitro We show that the N-terminal fragment of caspase-1-cleaved GSDMD rapidly targets the membrane fraction of macrophages and that it induces the formation of a plasma membrane pore. In vitro, the N-terminal fragment of caspase-1-cleaved recombinant GSDMD tightly binds liposomes and forms large permeability pores. Visualization of liposome-inserted GSDMD at nanometer resolution by cryo-electron and atomic force microscopy shows circular pores with variable ring diameters around 20 nm. Overall, these data demonstrate that GSDMD is the direct and final executor of pyroptotic cell death.

Abstract: Childhood stunting is the best overall indicator of children's well-being and an accurate reflection of social inequalities. Stunting is the most prevalent form of child malnutrition with an estimated 161 million children worldwide in 2013 falling below -2 SD from the length-for-age/height-for-age World Health Organization Child Growth Standards median. Many more millions suffer from some degree of growth faltering as the entire length-for-age/height-for-age z-score distribution is shifted to the left indicating that all children, and not only those falling below a specific cutoff, are affected. Despite global consensus on how to define and measure it, stunting often goes unrecognized in communities where short stature is the norm as linear growth is not routinely assessed in primary health care settings and it is difficult to visually recognize it. Growth faltering often begins in utero and continues for at least the first 2 years of post-natal life. Linear growth failure serves as a marker of multiple pathological disorders associated with increased morbidity and mortality, loss of physical growth potential, reduced neurodevelopmental and cognitive function and an elevated risk of chronic disease in adulthood. The severe irreversible physical and neurocognitive damage that accompanies stunted growth poses a major threat to human development. Increased awareness of stunting's magnitude and devastating consequences has resulted in its being identified as a major global health priority and the focus of international attention at the highest levels with global targets set for 2025 and beyond. The challenge is to prevent linear growth failure while keeping child overweight and obesity at bay.

Abstract: Estimating the size of the coronavirus disease 2019 (COVID-19) pandemic and the infection severity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is made challenging by inconsistencies in the available data. The number of deaths associated with COVID-19 is often used as a key indicator for the size of the epidemic, but the observed number of deaths represents only a minority of all infections1,2. In addition, the heterogeneous burdens in nursing homes and the variable reporting of deaths of older individuals can hinder direct comparisons of mortality rates and the underlying levels of transmission across countries3. Here we use age-specific COVID-19-associated death data from 45 countries and the results of 22 seroprevalence studies to investigate the consistency of infection and fatality patterns across multiple countries. We find that the age distribution of deaths in younger age groups (less than 65 years of age) is very consistent across different settings and demonstrate how these data can provide robust estimates of the share of the population that has been infected. We estimate that the infection fatality ratio is lowest among 5-9-year-old children, with a log-linear increase by age among individuals older than 30 years. Population age structures and heterogeneous burdens in nursing homes explain some but not all of the heterogeneity between countries in infection fatality ratios. Among the 45 countries included in our analysis, we estimate that approximately 5% of these populations had been infected by 1 September 2020, and that much higher transmission rates have probably occurred in a number of Latin American countries. This simple modelling framework can help countries to assess the progression of the pandemic and can be applied in any scenario for which reliable age-specific death data are available.

Abstract: The labor market increasingly rewards social skills. Between 1980 and 2012, jobs requiring high levels of social interaction grew by nearly 12 percentage points as a share of the U.S. labor force. Math-intensive but less social jobs—including many STEM occupations—shrank by 3.3 percentage points over the same period. Employment and wage growth were particularly strong for jobs requiring high levels of both math skill and social skills. To understand these patterns, I develop a model of team production where workers “trade tasks” to exploit their comparative advantage. In the model, social skills reduce coordination costs, allowing workers to specialize and work together more efficiently. The model generates predictions about sorting and the relative returns to skill across occupations, which I investigate using data from the NLSY79 and the NLSY97. Using a comparable set of skill measures and covariates across survey waves, I find that the labor market return to social skills was much greater in the 2000s than in the mid-1980s and 1990s.

Abstract: David E. Gray has produced a 4th edition of his successful textbook Doing Research in the Real World. This useful handbook for social science researchers includes the addition of entirely new chapters on Visual Research Methods, Digital Research Methods and Getting Started Using NVIVO, along with other chapters updated to reflect developments in digital research tools. This has been done whilst also retaining a comprehensive overview of research methods in the social sciences. The purpose of this book remains to assist researchers and their supervisors in navigating the complex decisions involved in the design and execution of research projects and does this in a clear and approachable way.

Abstract: Colorectal cancer (CRC) is the third most common cancer and the fourth most common cause of cancer-related death. Most cases of CRC are detected in Western countries, with its incidence increasing year by year. The probability of suffering from colorectal cancer is about 4%–5% and the risk for developing CRC is associated with personal features or habits such as age, chronic disease history and lifestyle. In this context, the gut microbiota has a relevant role, and dysbiosis situations can induce colonic carcinogenesis through a chronic inflammation mechanism. Some of the bacteria responsible for this multiphase process include Fusobacterium spp, Bacteroides fragilis and enteropathogenic Escherichia coli. CRC is caused by mutations that target oncogenes, tumour suppressor genes and genes related to DNA repair mechanisms. Depending on the origin of the mutation, colorectal carcinomas can be classified as sporadic (70%); inherited (5%) and familial (25%). The pathogenic mechanisms leading to this situation can be included in three types, namely chromosomal instability (CIN), microsatellite instability (MSI) and CpG island methylator phenotype (CIMP). Within these types of CRC, common mutations, chromosomal changes and translocations have been reported to affect important pathways (WNT, MAPK/PI3K, TGF-β, TP53), and mutations; in particular, genes such as c-MYC, KRAS, BRAF, PIK3CA, PTEN, SMAD2 and SMAD4 can be used as predictive markers for patient outcome. In addition to gene mutations, alterations in ncRNAs, such as lncRNA or miRNA, can also contribute to different steps of the carcinogenesis process and have a predictive value when used as biomarkers. In consequence, different panels of genes and mRNA are being developed to improve prognosis and treatment selection. The choice of first-line treatment in CRC follows a multimodal approach based on tumour-related characteristics and usually comprises surgical resection followed by chemotherapy combined with monoclonal antibodies or proteins against vascular endothelial growth factor (VEGF) and epidermal growth receptor (EGFR). Besides traditional chemotherapy, alternative therapies (such as agarose tumour macrobeads, anti-inflammatory drugs, probiotics, and gold-based drugs) are currently being studied to increase treatment effectiveness and reduce side effects.

Abstract: Obesity is an increasingly prevalent disease regulated by genetic and environmental factors. Emerging studies indicate that immune cells, including monocytes, granulocytes and lymphocytes, regulate metabolic homeostasis and are dysregulated in obesity. Group 2 innate lymphoid cells (ILC2s) can regulate adaptive immunity and eosinophil and alternatively activated macrophage responses, and were recently identified in murine white adipose tissue (WAT) where they may act to limit the development of obesity. However, ILC2s have not been identified in human adipose tissue, and the mechanisms by which ILC2s regulate metabolic homeostasis remain unknown. Here we identify ILC2s in human WAT and demonstrate that decreased ILC2 responses in WAT are a conserved characteristic of obesity in humans and mice. Interleukin (IL)-33 was found to be critical for the maintenance of ILC2s in WAT and in limiting adiposity in mice by increasing caloric expenditure. This was associated with recruitment of uncoupling protein 1 (UCP1)(+) beige adipocytes in WAT, a process known as beiging or browning that regulates caloric expenditure. IL-33-induced beiging was dependent on ILC2s, and IL-33 treatment or transfer of IL-33-elicited ILC2s was sufficient to drive beiging independently of the adaptive immune system, eosinophils or IL-4 receptor signalling. We found that ILC2s produce methionine-enkephalin peptides that can act directly on adipocytes to upregulate Ucp1 expression in vitro and that promote beiging in vivo. Collectively, these studies indicate that, in addition to responding to infection or tissue damage, ILC2s can regulate adipose function and metabolic homeostasis in part via production of enkephalin peptides that elicit beiging.

Abstract: We propose a class of loss functions, which we call deep perceptual similarity metrics (DeePSiM), allowing to generate sharp high resolution images from compressed abstract representations. Instead of computing distances in the image space, we compute distances between image features extracted by deep neural networks. This metric reflects perceptual similarity of images much better and, thus, leads to better results. We demonstrate two examples of use cases of the proposed loss: (1) networks that invert the AlexNet convolutional network; (2) a modified version of a variational autoencoder that generates realistic high-resolution random images.

Abstract: The current strive towards end-to-end trainable computer vision systems imposes major challenges for the task of visual tracking. In contrast to most other vision problems, tracking requires the learning of a robust target-specific appearance model online, during the inference stage. To be end-to-end trainable, the online learning of the target model thus needs to be embedded in the tracking architecture itself. Due to the imposed challenges, the popular Siamese paradigm simply predicts a target feature template, while ignoring the background appearance information during inference. Consequently, the predicted model possesses limited target-background discriminability. We develop an end-to-end tracking architecture, capable of fully exploiting both target and background appearance information for target model prediction. Our architecture is derived from a discriminative learning loss by designing a dedicated optimization process that is capable of predicting a powerful model in only a few iterations. Furthermore, our approach is able to learn key aspects of the discriminative loss itself. The proposed tracker sets a new state-of-the-art on 6 tracking benchmarks, achieving an EAO score of 0.440 on VOT2018, while running at over 40 FPS. The code and models are available at https://github.com/visionml/pytracking.

Abstract: We address temporal action localization in untrimmed long videos. This is important because videos in real applications are usually unconstrained and contain multiple action instances plus video content of background scenes or other activities. To address this challenging issue, we exploit the effectiveness of deep networks in temporal action localization via three segment-based 3D ConvNets: (1) a proposal network identifies candidate segments in a long video that may contain actions, (2) a classification network learns one-vs-all action classification model to serve as initialization for the localization network, and (3) a localization network fine-tunes the learned classification network to localize each action instance. We propose a novel loss function for the localization network to explicitly consider temporal overlap and achieve high temporal localization accuracy. In the end, only the proposal network and the localization network are used during prediction. On two largescale benchmarks, our approach achieves significantly superior performances compared with other state-of-the-art systems: mAP increases from 1.7% to 7.4% on MEXaction2 and increases from 15.0% to 19.0% on THUMOS 2014.

Abstract: El Nino–Southern Oscillation (ENSO) is a naturally occurring mode of tropical Pacific variability, with global impacts on society and natural ecosystems. While it has long been known that El Nino events display a diverse range of amplitudes, triggers, spatial patterns, and life cycles, the realization that ENSO’s impacts can be highly sensitive to this event-to-event diversity is driving a renewed interest in the subject. This paper surveys our current state of knowledge of ENSO diversity, identifies key gaps in understanding, and outlines some promising future research directions.

Abstract: Cryptocurrencies, based on and led by Bitcoin, have shown promise as infrastructure for pseudonymous online payments, cheap remittance, trustless digital asset exchange, and smart contracts. However, Bitcoin-derived blockchain protocols have inherent scalability limits that trade-off between throughput and latency and withhold the realization of this potential. This paper presents Bitcoin-NG, a new blockchain protocol designed to scale. Based on Bitcoin's blockchain protocol, Bitcoin-NG is Byzantine fault tolerant, is robust to extreme churn, and shares the same trust model obviating qualitative changes to the ecosystem. In addition to Bitcoin-NG, we introduce several novel metrics of interest in quantifying the security and efficiency of Bitcoin-like blockchain protocols. We implement Bitcoin-NG and perform large-scale experiments at 15% the size of the operational Bitcoin system, using unchanged clients of both protocols. These experiments demonstrate that Bitcoin-NG scales optimally, with bandwidth limited only by the capacity of the individual nodes and latency limited only by the propagation time of the network.

Abstract: We develop a conceptual framework that allows us to define the sharing economy and its close cousins and we understand its sudden rise from an economic-historic perspective. We then assess the sharing economy platforms in terms of the economic, social and environmental impacts. We end with reflections on current regulations and future alternatives, and suggest a number of future research questions.

Abstract: The Cloud infrastructure and its extensive set of Internet-accessible resources has potential to provide significant benefits to robots and automation systems. We consider robots and automation systems that rely on data or code from a network to support their operation, i.e., where not all sensing, computation, and memory is integrated into a standalone system. This survey is organized around four potential benefits of the Cloud: 1) Big Data: access to libraries of images, maps, trajectories, and descriptive data; 2) Cloud Computing: access to parallel grid computing on demand for statistical analysis, learning, and motion planning; 3) Collective Robot Learning: robots sharing trajectories, control policies, and outcomes; and 4) Human Computation: use of crowdsourcing to tap human skills for analyzing images and video, classification, learning, and error recovery. The Cloud can also improve robots and automation systems by providing access to: a) datasets, publications, models, benchmarks, and simulation tools; b) open competitions for designs and systems; and c) open-source software. This survey includes over 150 references on results and open challenges. A website with new developments and updates is available at: http://goldberg.berkeley.edu/cloud-robotics/

Abstract: AutoDock Vina is arguably one of the fastest and most widely used open-source programs for molecular docking. However, compared to other programs in the AutoDock Suite, it lacks support for modeling specific features such as macrocycles or explicit water molecules. Here, we describe the implementation of this functionality in AutoDock Vina 1.2.0. Additionally, AutoDock Vina 1.2.0 supports the AutoDock4.2 scoring function, simultaneous docking of multiple ligands, and a batch mode for docking a large number of ligands. Furthermore, we implemented Python bindings to facilitate scripting and the development of docking workflows. This work is an effort toward the unification of the features of the AutoDock4 and AutoDock Vina programs. The source code is available at https://github.com/ccsb-scripps/AutoDock-Vina.

Abstract: To explore the origins and consequences of tetraploidy in the African clawed frog, we sequenced the Xenopus laevis genome and compared it to the related diploid X. tropicalis genome. We characterize the allotetraploid origin of X. laevis by partitioning its genome into two homoeologous subgenomes, marked by distinct families of 'fossil' transposable elements. On the basis of the activity of these elements and the age of hundreds of unitary pseudogenes, we estimate that the two diploid progenitor species diverged around 34 million years ago (Ma) and combined to form an allotetraploid around 17-18 Ma. More than 56% of all genes were retained in two homoeologous copies. Protein function, gene expression, and the amount of conserved flanking sequence all correlate with retention rates. The subgenomes have evolved asymmetrically, with one chromosome set more often preserving the ancestral state and the other experiencing more gene loss, deletion, rearrangement, and reduced gene expression.

Abstract: We present the second Gaia data release, Gaia DR2, consisting of astrometry, photometry, radial velocities, and information on astrophysical parameters and variability, for sources brighter than magnitude 21. In addition epoch astrometry and photometry are provided for a modest sample of minor planets in the solar system. A summary of the contents of Gaia DR2 is presented, accompanied by a discussion on the differences with respect to Gaia DR1 and an overview of the main limitations which are still present in the survey. Recommendations are made on the responsible use of Gaia DR2 results. Gaia DR2 contains celestial positions and the apparent brightness in G for approximately 1.7 billion sources. For 1.3 billion of those sources, parallaxes and proper motions are in addition available. The sample of sources for which variability information is provided is expanded to 0.5 million stars. This data release contains four new elements: broad-band colour information in the form of the apparent brightness in the $G_\mathrm{BP}$ (330--680 nm) and $G_\mathrm{RP}$ (630--1050 nm) bands is available for 1.4 billion sources; median radial velocities for some 7 million sources are presented; for between 77 and 161 million sources estimates are provided of the stellar effective temperature, extinction, reddening, and radius and luminosity; and for a pre-selected list of 14000 minor planets in the solar system epoch astrometry and photometry are presented. Finally, Gaia DR2 also represents a new materialisation of the celestial reference frame in the optical, the Gaia-CRF2, which is the first optical reference frame based solely on extragalactic sources. There are notable changes in the photometric system and the catalogue source list with respect to Gaia DR1, and we stress the need to consider the two data releases as independent.