Showing papers by "University of Luxembourg published in 2017"

Quantum-chemical insights from deep tensor neural networks.

Abstract: Learning from data has led to paradigm shifts in a multitude of disciplines, including web, text and image search, speech recognition, as well as bioinformatics. Can machine learning enable similar breakthroughs in understanding quantum many-body systems? Here we develop an efficient deep learning approach that enables spatially and chemically resolved insights into quantum-mechanical observables of molecular systems. We unify concepts from many-body Hamiltonians with purpose-designed deep tensor neural networks, which leads to size-extensive and uniformly accurate (1 kcal mol−1) predictions in compositional and configurational chemical space for molecules of intermediate size. As an example of chemical relevance, the model reveals a classification of aromatic rings with respect to their stability. Further applications of our model for predicting atomic energies and local chemical potentials in molecules, reliable isomer energies, and molecules with peculiar electronic structure demonstrate the potential of machine learning for revealing insights into complex quantum-chemical systems. Machine learning is an increasingly popular approach to analyse data and make predictions. Here the authors develop a ‘deep learning’ framework for quantitative predictions and qualitative understanding of quantum-mechanical observables of chemical systems, beyond properties trivially contained in the training data.

Machine learning of accurate energy-conserving molecular force fields

Abstract: Using conservation of energy-a fundamental property of closed classical and quantum mechanical systems-we develop an efficient gradient-domain machine learning (GDML) approach to construct accurate molecular force fields using a restricted number of samples from ab initio molecular dynamics (AIMD) trajectories. The GDML implementation is able to reproduce global potential energy surfaces of intermediate-sized molecules with an accuracy of 0.3 kcal mol-1 for energies and 1 kcal mol-1 A-1 for atomic forces using only 1000 conformational geometries for training. We demonstrate this accuracy for AIMD trajectories of molecules, including benzene, toluene, naphthalene, ethanol, uracil, and aspirin. The challenge of constructing conservative force fields is accomplished in our work by learning in a Hilbert space of vector-valued functions that obey the law of energy conservation. The GDML approach enables quantitative molecular dynamics simulations for molecules at a fraction of cost of explicit AIMD calculations, thereby allowing the construction of efficient force fields with the accuracy and transferability of high-level ab initio methods.

The Human Phenotype Ontology in 2017

Abstract: Deep phenotyping has been defined as the precise and comprehensive analysis of phenotypic abnormalities in which the individual components of the phenotype are observed and described. The three components of the Human Phenotype Ontology (HPO; www.human-phenotype-ontology.org) project are the phenotype vocabulary, disease-phenotype annotations and the algorithms that operate on these. These components are being used for computational deep phenotyping and precision medicine as well as integration of clinical data into translational research. The HPO is being increasingly adopted as a standard for phenotypic abnormalities by diverse groups such as international rare disease organizations, registries, clinical labs, biomedical resources, and clinical software tools and will thereby contribute toward nascent efforts at global data exchange for identifying disease etiologies. This update article reviews the progress of the HPO project since the debut Nucleic Acids Research database article in 2014, including specific areas of expansion such as common (complex) disease, new algorithms for phenotype driven genomic discovery and diagnostics, integration of cross-species mapping efforts with the Mammalian Phenotype Ontology, an improved quality control pipeline, and the addition of patient-friendly terminology.

SchNet - a deep learning architecture for molecules and materials

Abstract: Deep learning has led to a paradigm shift in artificial intelligence, including web, text and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning in general and deep learning in particular is ideally suited for representing quantum-mechanical interactions, enabling to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for \emph{molecules and materials} where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study of the quantum-mechanical properties of C$_{20}$-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Generation of genome-scale metabolic reconstructions for 773 members of the human gut microbiota

Abstract: A large set of microbial metabolic models (AGORA) could be applied to better understand the functions of the human gut microbiome. Genome-scale metabolic models derived from human gut metagenomic data can be used as a framework to elucidate how microbial communities modulate human metabolism and health. We present AGORA (assembly of gut organisms through reconstruction and analysis), a resource of genome-scale metabolic reconstructions semi-automatically generated for 773 human gut bacteria. Using this resource, we identified a defined growth medium for Bacteroides caccae ATCC 34185. We also showed that interactions among modeled species depend on both the metabolic potential of each species and the nutrients available. AGORA reconstructions can integrate either metagenomic or 16S rRNA sequencing data sets to infer the metabolic diversity of microbial communities. AGORA reconstructions could provide a starting point for the generation of high-quality, manually curated metabolic reconstructions. AGORA is fully compatible with Recon 2, a comprehensive metabolic reconstruction of human metabolism, which will facilitate studies of host–microbiome interactions.

Dopamine oxidation mediates mitochondrial and lysosomal dysfunction in Parkinson’s disease

Abstract: Mitochondrial and lysosomal dysfunction have been implicated in substantia nigra dopaminergic neurodegeneration in Parkinson’s disease (PD), but how these pathways are linked in human neurons remains unclear. Here we studied dopaminergic neurons derived from patients with idiopathic and familial PD. We identified a time-dependent pathological cascade beginning with mitochondrial oxidant stress leading to oxidized dopamine accumulation and ultimately resulting in reduced glucocerebrosidase enzymatic activity, lysosomal dysfunction, and α-synuclein accumulation. This toxic cascade was observed in human, but not in mouse, PD neurons at least in part because of species-specific differences in dopamine metabolism. Increasing dopamine synthesis or α-synuclein amounts in mouse midbrain neurons recapitulated pathological phenotypes observed in human neurons. Thus, dopamine oxidation represents an important link between mitochondrial and lysosomal dysfunction in PD pathogenesis.

First-Principles Models for van der Waals Interactions in Molecules and Materials: Concepts, Theory, and Applications.

Abstract: Noncovalent van der Waals (vdW) or dispersion forces are ubiquitous in nature and influence the structure, stability, dynamics, and function of molecules and materials throughout chemistry, biology, physics, and materials science. These forces are quantum mechanical in origin and arise from electrostatic interactions between fluctuations in the electronic charge density. Here, we explore the conceptual and mathematical ingredients required for an exact treatment of vdW interactions, and present a systematic and unified framework for classifying the current first-principles vdW methods based on the adiabatic-connection fluctuation–dissipation (ACFD) theorem (namely the Rutgers–Chalmers vdW-DF, Vydrov–Van Voorhis (VV), exchange-hole dipole moment (XDM), Tkatchenko–Scheffler (TS), many-body dispersion (MBD), and random-phase approximation (RPA) approaches). Particular attention is paid to the intriguing nature of many-body vdW interactions, whose fundamental relevance has recently been highlighted in several...

Learning through ferroelectric domain dynamics in solid-state synapses.

Abstract: In the brain, learning is achieved through the ability of synapses to reconfigure the strength by which they connect neurons (synaptic plasticity). In promising solid-state synapses called memristors, conductance can be finely tuned by voltage pulses and set to evolve according to a biological learning rule called spike-timing-dependent plasticity (STDP). Future neuromorphic architectures will comprise billions of such nanosynapses, which require a clear understanding of the physical mechanisms responsible for plasticity. Here we report on synapses based on ferroelectric tunnel junctions and show that STDP can be harnessed from inhomogeneous polarization switching. Through combined scanning probe imaging, electrical transport and atomic-scale molecular dynamics, we demonstrate that conductance variations can be modelled by the nucleation-dominated reversal of domains. Based on this physical model, our simulations show that arrays of ferroelectric nanosynapses can autonomously learn to recognize patterns in a predictable way, opening the path towards unsupervised learning in spiking neural networks.

Derivation of human midbrain-specific organoids from neuroepithelial Stem Cells

Abstract: Research on human brain development and neurological diseases is limited by the lack of advanced experimental in vitro models that truly recapitulate the complexity of the human brain. Here, we describe a robust human brain organoid system that is highly specific to the midbrain derived from regionally patterned neuroepithelial stem cells. These human midbrain organoids contain spatially organized groups of dopaminergic neurons, which make them an attractive model for the study of Parkinson’s disease. Midbrain organoids are characterized in detail for neuronal, astroglial, and oligodendrocyte differentiation. Furthermore, we show the presence of synaptic connections and electrophysiological activity. The complexity of this model is further highlighted by the myelination of neurites. The present midbrain organoid system has the potential to be used for advanced in vitro disease modeling and therapy development.

Quantum and Information Thermodynamics: A Unifying Framework Based on Repeated Interactions

Abstract: Nanomachines are subject to random thermal and quantum fluctuations that are not captured by traditional thermodynamic theory. A new theoretical investigation offers a step toward a unified nanoscale theory by showing how externally prepared systems (e.g., atoms in an optical cavity or DNA bases in an enzyme reaction) that interact with a nanoscopic device can be a source of nonequilbrium free energy.

Static analysis of android apps

Abstract: ContextStatic analysis exploits techniques that parse program source code or bytecode, often traversing program paths to check some program properties. Static analysis approaches have been proposed for different tasks, including for assessing the security of Android apps, detecting app clones, automating test cases generation, or for uncovering non-functional issues related to performance or energy. The literature thus has proposed a large body of works, each of which attempts to tackle one or more of the several challenges that program analyzers face when dealing with Android apps. ObjectiveWe aim to provide a clear view of the state-of-the-art works that statically analyze Android apps, from which we highlight the trends of static analysis approaches, pinpoint where the focus has been put, and enumerate the key aspects where future researches are still needed. MethodWe have performed a systematic literature review (SLR) which involves studying 124 research papers published in software engineering, programming languages and security venues in the last 5 years (January 2011December 2015). This review is performed mainly in five dimensions: problems targeted by the approach, fundamental techniques used by authors, static analysis sensitivities considered, android characteristics taken into account and the scale of evaluation performed. ResultsOur in-depth examination has led to several key findings: 1) Static analysis is largely performed to uncover security and privacy issues; 2) The Soot framework and the Jimple intermediate representation are the most adopted basic support tool and format, respectively; 3) Taint analysis remains the most applied technique in research approaches; 4) Most approaches support several analysis sensitivities, but very few approaches consider path-sensitivity; 5) There is no single work that has been proposed to tackle all challenges of static analysis that are related to Android programming; and 6) Only a small portion of state-of-the-art works have made their artifacts publicly available. ConclusionThe research community is still facing a number of challenges for building approaches that are aware altogether of implicit-Flows, dynamic code loading features, reflective calls, native code and multi-threading, in order to implement sound and highly precise static analyzers.

SchNet: A continuous-filter convolutional neural network for modeling quantum interactions

Abstract: Deep learning has the potential to revolutionize quantum chemistry as it is ideally suited to learn representations for structured data and speed up the exploration of chemical space. While convolutional neural networks have proven to be the first choice for images, audio and video data, the atoms in molecules are not restricted to a grid. Instead, their precise locations contain essential physical information, that would get lost if discretized. Thus, we propose to use continuous-filter convolutional layers to be able to model local correlations without requiring the data to lie on a grid. We apply those layers in SchNet: a novel deep learning architecture modeling quantum interactions in molecules. We obtain a joint model for the total energy and interatomic forces that follows fundamental quantum-chemical principles. Our architecture achieves state-of-the-art performance for benchmarks of equilibrium molecules and molecular dynamics trajectories. Finally, we introduce a more challenging benchmark with chemical and structural variations that suggests the path for further work.

BCAT1 restricts αKG levels in AML stem cells leading to IDHmut-like DNA hypermethylation

Abstract: The branched-chain amino acid (BCAA) pathway and high levels of BCAA transaminase 1 (BCAT1) have recently been associated with aggressiveness in several cancer entities. However, the mechanistic role of BCAT1 in this process remains largely uncertain. Here, by performing high-resolution proteomic analysis of human acute myeloid leukaemia (AML) stem-cell and non-stem-cell populations, we find the BCAA pathway enriched and BCAT1 protein and transcripts overexpressed in leukaemia stem cells. We show that BCAT1, which transfers α-amino groups from BCAAs to α-ketoglutarate (αKG), is a critical regulator of intracellular αKG homeostasis. Further to its role in the tricarboxylic acid cycle, αKG is an essential cofactor for αKG-dependent dioxygenases such as Egl-9 family hypoxia inducible factor 1 (EGLN1) and the ten-eleven translocation (TET) family of DNA demethylases. Knockdown of BCAT1 in leukaemia cells caused accumulation of αKG, leading to EGLN1-mediated HIF1α protein degradation. This resulted in a growth and survival defect and abrogated leukaemia-initiating potential. By contrast, overexpression of BCAT1 in leukaemia cells decreased intracellular αKG levels and caused DNA hypermethylation through altered TET activity. AML with high levels of BCAT1 (BCAT1high) displayed a DNA hypermethylation phenotype similar to cases carrying a mutant isocitrate dehydrogenase (IDHmut), in which TET2 is inhibited by the oncometabolite 2-hydroxyglutarate. High levels of BCAT1 strongly correlate with shorter overall survival in IDHWTTET2WT, but not IDHmut or TET2mut AML. Gene sets characteristic for IDHmut AML were enriched in samples from patients with an IDHWTTET2WTBCAT1high status. BCAT1high AML showed robust enrichment for leukaemia stem-cell signatures, and paired sample analysis showed a significant increase in BCAT1 levels upon disease relapse. In summary, by limiting intracellular αKG, BCAT1 links BCAA catabolism to HIF1α stability and regulation of the epigenomic landscape, mimicking the effects of IDH mutations. Our results suggest the BCAA-BCAT1-αKG pathway as a therapeutic target to compromise leukaemia stem-cell function in patients with IDHWTTET2WT AML.

SchNet: A continuous-filter convolutional neural network for modeling quantum interactions

Abstract: Deep learning has the potential to revolutionize quantum chemistry as it is ideally suited to learn representations for structured data and speed up the exploration of chemical space. While convolutional neural networks have proven to be the first choice for images, audio and video data, the atoms in molecules are not restricted to a grid. Instead, their precise locations contain essential physical information, that would get lost if discretized. Thus, we propose to use continuous-filter convolutional layers to be able to model local correlations without requiring the data to lie on a grid. We apply those layers in SchNet: a novel deep learning architecture modeling quantum interactions in molecules. We obtain a joint model for the total energy and interatomic forces that follows fundamental quantum-chemical principles. This includes rotationally invariant energy predictions and a smooth, differentiable potential energy surface. Our architecture achieves state-of-the-art performance for benchmarks of equilibrium molecules and molecular dynamics trajectories. Finally, we introduce a more challenging benchmark with chemical and structural variations that suggests the path for further work.

Colonization and Succession within the Human Gut Microbiome by Archaea, Bacteria, and Microeukaryotes during the First Year of Life

Abstract: Perturbations to the colonization process of the human gastrointestinal tract have been suggested to result in adverse health effects later in life. Although much research has been performed on bacterial colonization and succession, much less is known about the other two domains of life, archaea, and eukaryotes. Here we describe colonization and succession by bacteria, archaea and microeukaryotes during the first year of life (samples collected around days 1, 3, 5, 28, 150, and 365) within the gastrointestinal tract of infants delivered either vaginally or by cesarean section and using a combination of quantitative real-time PCR as well as 16S and 18S rRNA gene amplicon sequencing. Sequences from organisms belonging to all three domains of life were detectable in all of the collected meconium samples. The microeukaryotic community composition fluctuated strongly over time and early diversification was delayed in infants receiving formula milk. Cesarean section-delivered (CSD) infants experienced a delay in colonization and succession, which was observed for all three domains of life. Shifts in prokaryotic succession in CSD infants compared to vaginally delivered (VD) infants were apparent as early as days 3 and 5, which were characterized by increased relative abundances of the genera Streptococcus and Staphylococcus, and a decrease in relative abundance for the genera Bifidobacterium and Bacteroides. Generally, a depletion in Bacteroidetes was detected as early as day 5 postpartum in CSD infants, causing a significantly increased Firmicutes/Bacteroidetes ratio between days 5 and 150 when compared to VD infants. Although the delivery mode appeared to have the strongest influence on differences between the infants, other factors such as a younger gestational age or maternal antibiotics intake likely contributed to the observed patterns as well. Our findings complement previous observations of a delay in colonization and succession of CSD infants, which affects not only bacteria but also archaea and microeukaryotes. This further highlights the need for resolving bacterial, archaeal, and microeukaryotic dynamics in future longitudinal studies of microbial colonization and succession within the neonatal gastrointestinal tract.

Long Noncoding RNA MANTIS Facilitates Endothelial Angiogenic Function

Abstract: Background:The angiogenic function of endothelial cells is regulated by numerous mechanisms, but the impact of long noncoding RNAs (lncRNAs) has hardly been studied. We set out to identify novel an...

BacArena: Individual-based metabolic modeling of heterogeneous microbes in complex communities

Abstract: Recent advances focusing on the metabolic interactions within and between cellular populations have emphasized the importance of microbial communities for human health Constraint-based modeling, with flux balance analysis in particular, has been established as a key approach for studying microbial metabolism, whereas individual-based modeling has been commonly used to study complex dynamics between interacting organisms In this study, we combine both techniques into the R package BacArena (https://cranr-projectorg/package=BacArena) to generate novel biological insights into Pseudomonas aeruginosa biofilm formation as well as a seven species model community of the human gut For our P aeruginosa model, we found that cross-feeding of fermentation products cause a spatial differentiation of emerging metabolic phenotypes in the biofilm over time In the human gut model community, we found that spatial gradients of mucus glycans are important for niche formations which shape the overall community structure Additionally, we could provide novel hypothesis concerning the metabolic interactions between the microbes These results demonstrate the importance of spatial and temporal multi-scale modeling approaches such as BacArena

Self-reported dependence on mobile phones in young adults: a European cross-cultural empirical survey

Abstract: Background and aims Despite many positive benefits, mobile phone use can be associated with harmful and detrimental behaviors. The aim of this study was twofold: to examine (a) cross-cultural patterns of perceived dependence on mobile phones in ten European countries, first, grouped in four different regions (North: Finland and UK; South: Spain and Italy; East: Hungary and Poland; West: France, Belgium, Germany, and Switzerland), and second by country, and (b) how socio-demographics, geographic differences, mobile phone usage patterns, and associated activities predicted this perceived dependence. Methods A sample of 2,775 young adults (aged 18-29 years) were recruited in different European Universities who participated in an online survey. Measures included socio-demographic variables, patterns of mobile phone use, and the dependence subscale of a short version of the Problematic Mobile Phone Use Questionnaire (PMPUQ; Billieux, Van der Linden, & Rochat, 2008). Results The young adults from the Northern and Southern regions reported the heaviest use of mobile phones, whereas perceived dependence was less prevalent in the Eastern region. However, the proportion of highly dependent mobile phone users was more elevated in Belgium, UK, and France. Regression analysis identified several risk factors for increased scores on the PMPUQ dependence subscale, namely using mobile phones daily, being female, engaging in social networking, playing video games, shopping and viewing TV shows through the Internet, chatting and messaging, and using mobile phones for downloading-related activities. Discussion and conclusions Self-reported dependence on mobile phone use is influenced by frequency and specific application usage.

The Challenge of Non-Technical Loss Detection using Artificial Intelligence: A Survey

Abstract: Detection of non-technical losses (NTL) which include electricity theft, faulty meters or billing errors has attracted increasing attention from researchers in electrical engineering and computer science. NTLs cause significant harm to the economy, as in some countries they may range up to 40% of the total electricity distributed. The predominant research direction is employing artificial intelligence to predict whether a customer causes NTL. This paper first provides an overview of how NTLs are defined and their impact on economies, which include loss of revenue and profit of electricity providers and decrease of the stability and reliability of electrical power grids. It then surveys the state-of-the-art research efforts in a up-to-date and comprehensive review of algorithms, features and data sets used. It finally identifies the key scientific and engineering challenges in NTL detection and suggests how they could be addressed in the future.

Hybrid software and system development in practice: waterfall, scrum, and beyond

Abstract: Software and system development faces numerous challenges of rapidly changing markets. To address such challenges, companies and projects design and adopt specific development approaches by combining well-structured comprehensive methods and flexible agile practices. Yet, the number of methods and practices is large, and available studies argue that the actual process composition is carried out in a fairly ad-hoc manner. The present paper reports on a survey on hybrid software development approaches. We study which approaches are used in practice, how different approaches are combined, and what contextual factors influence the use and combination of hybrid software development approaches. Our results from 69 study participants show a variety of development approaches used and combined in practice. We show that most combinations follow a pattern in which a traditional process model serves as framework in which several fine-grained (agile) practices are plugged in. We further show that hybrid software development approaches are independent from the company size and external triggers. We conclude that such approaches are the results of a natural process evolution, which is mainly driven by experience, learning, and pragmatism.

Codebook-Based Hybrid Precoding for Millimeter Wave Multiuser Systems

Abstract: In millimeter-wave (mmWave) systems, antenna architecture limitations make it difficult to apply conventional fully digital precoding techniques but call for low-cost analog radio frequency (RF) and digital baseband hybrid precoding methods. This paper investigates joint RF-baseband hybrid precoding for the downlink of multiuser multiantenna mmWave systems with a limited number of RF chains. Two performance measures, maximizing the spectral efficiency and the energy efficiency of the system, are considered. We propose a codebook-based RF precoding design and obtain the channel state information via a beam sweep procedure. Via the codebook-based design, the original system is transformed into a virtual multiuser downlink system with the RF chain constraint. Consequently, we are able to simplify the complicated hybrid precoding optimization problems to joint codeword selection and precoder design (JWSPD) problems. Then, we propose efficient methods to address the JWSPD problems and jointly optimize the RF and baseband precoders under the two performance measures. Finally, extensive numerical results are provided to validate the effectiveness of the proposed hybrid precoders.

Variation in GP decisions on antihypertensive treatment in oldest-old and frail individuals across 29 countries

Abstract: In oldest-old patients (>80), few trials showed efficacy of treating hypertension and they included mostly the healthiest elderly. The resulting lack of knowledge has led to inconsistent guidelines, mainly based on systolic blood pressure (SBP), cardiovascular disease (CVD) but not on frailty despite the high prevalence in oldest-old. This may lead to variation how General Practitioners (GPs) treat hypertension. Our aim was to investigate treatment variation of GPs in oldest-olds across countries and to identify the role of frailty in that decision. Using a survey, we compared treatment decisions in cases of oldest-old varying in SBP, CVD, and frailty. GPs were asked if they would start antihypertensive treatment in each case. In 2016, we invited GPs in Europe, Brazil, Israel, and New Zealand. We compared the percentage of cases that would be treated per countries. A logistic mixed-effects model was used to derive odds ratio (OR) for frailty with 95% confidence intervals (CI), adjusted for SBP, CVD, and GP characteristics (sex, location and prevalence of oldest-old per GP office, and years of experience). The mixed-effects model was used to account for the multiple assessments per GP. The 29 countries yielded 2543 participating GPs: 52% were female, 51% located in a city, 71% reported a high prevalence of oldest-old in their offices, 38% and had >20 years of experience. Across countries, considerable variation was found in the decision to start antihypertensive treatment in the oldest-old ranging from 34 to 88%. In 24/29 (83%) countries, frailty was associated with GPs’ decision not to start treatment even after adjustment for SBP, CVD, and GP characteristics (OR 0.53, 95%CI 0.48–0.59; ORs per country 0.11–1.78). Across countries, we found considerable variation in starting antihypertensive medication in oldest-old. The frail oldest-old had an odds ratio of 0.53 of receiving antihypertensive treatment. Future hypertension trials should also include frail patients to acquire evidence on the efficacy of antihypertensive treatment in oldest-old patients with frailty, with the aim to get evidence-based data for clinical decision-making.

The ICO Gold Rush: It's a Scam, It's a Bubble, It's a Super Challenge for Regulators

Abstract: Initial coin offerings typically use blockchain technology to offer tokens that confer various rights in return, most often, for cryptocurrency. They can be seen as a conjunction of crowdfunding and blockchain. Based on a database of over 1000 ICO white papers, we provide a taxonomy of ICOs to increase understanding of their many forms, analyze the various regulatory challenges they pose, and suggest the first steps regulators should consider in response. As our database shows, ICOs are a global phenomenon and the global ICO market is much larger than generally thought, with overall ICO subscriptions estimated to exceed 75 billion USD as at the end of June 2. The US ICO market is significant, but the US doesn’t dominate this market, by any means. Furthermore, many ICOs are offered on the basis of utterly inadequate disclosure of information; more than half the ICO white papers are either silent on the initiators or backers or do not provide contact details, and an even greater share do not elaborate on the applicable law, segregation or pooling of client funds, and the existence of an external auditor. Accordingly, the decision to invest in them often cannot be the outcome of a rational calculus. Hallmarks of a classic speculative bubble are present. At the same time, ICOs provide a new, innovative and potentially important vehicle for raising funds to support innovative ideas and ventures, with the potential for aspects of their underlying structure to have an important impact on fundraising systems and structures in future.