KITopen
Not a member yet
63711 research outputs found
Sort by
Chromatin‐associated condensates as an inspiration for the system architecture of future DNA computers
Get PDFThe genome stores and processes approximately 1.5 gigabytes of encoded information. In this article, we propose that the eukaryotic genome and its adaptable three-dimensional packing in the form of chromatin offer a valuable template for the system architecture of DNA-based digital computers. We examine embryonic and stem cells, which exhibit distinct chromatin-associated condensates enriched in transcription machinery. These dynamic biomolecular condensates facilitate the spatial association of genes, genomic control elements, and molecular machinery responsible for reading the genomic code. Drawing a compelling analogy to the von Neumann computer architecture-which integrates storage, processing, and memory in most electronic computers-we reflect on how the operational principles of these condensates could inspire the design of a similar architecture for future DNA computers. In particular, we describe how one could recreate such an architecture by exploiting the process of surface condensation, which underlies the formation of chromatin-associated condensates. We conclude by reviewing our initial steps of constructing synthetic DNA nanostructures that follow the same operational principles and enable programmable surface condensation. Finally, we outline how computational methods from accelerated materials design could further advance the development of DNA computer system architectures
Search for pair production of heavy particles decaying to a top quark and a gluon in the lepton+jets final state in proton–proton collisions at
Get PDFA search is presented for the pair production of new heavy resonances, each decaying into a top quark (t) or antiquark and a gluon (g). The analysis uses data recorded with the CMS detector from proton-proton collisions at a center-of-mass energy of 13 TeV at the LHC, corresponding to an integrated luminosity of 138 fb. Events with one muon or electron, multiple jets, and missing transverse momentum are selected. After using a deep neural network to enrich the data sample with signal-like events, distributions in the scalar sum of the transverse momenta of all reconstructed objects are analyzed in the search for a signal. No significant deviations from the standard model prediction are found. Upper limits at 95% confidence level are set on the product of cross section and branching fraction squared for the pair production of excited top quarks in the t → tg decay channel. The upper limits range from 120 to 0.8 fb for
a t with spin-1/2 and from 15 to 1.0 fb for a t∗ with spin-3/2. These correspond to mass exclusion limits up to 1050 and 1700 GeV for spin-1/2 and spin-3/2 t particles, respectively. These are the most stringent limits to date on the existence of t → tg resonances
Search for excited tau leptons in the ττγ final state in proton-proton collisions at = 13 TeV
Get PDFResults are presented for a test of the compositeness of the heaviest charged lepton, τ, using data collected by the CMS experiment in proton-proton collisions at a center-of-mass energy of 13 TeV at the CERN LHC. The data were collected in 2016-2018 and correspond to an integrated luminosity of 138 fb. This analysis searches for tau lepton pair production in which one of the tau leptons is produced in an excited state and decays to a ground state tau lepton and a photon. The event selection consists of two isolated tau lepton decay candidates and a high-energy photon. The mass of the excited tau lepton is reconstructed using the missing transverse momentum in the event, assuming the momentum of the neutrinos from each tau lepton decay are aligned with the visible decay products. No excess of events above the standard model background prediction is observed. This null result is used to set lower bounds on the excited tau lepton mass. For a compositeness scale equal to the excited tau lepton mass, excited tau leptons with masses below 4700 GeV are excluded at 95% confidence level; for = 10 TeV this exclusion is set at 2800 GeV. This is the first experimental result covering this production and decay process in the excited tau mass range above 175 GeV
Bifurcation in Stick–Slip-Induced Low-Frequency Brake Noises: Experimental and Numerical Study
Get PDFThe term honk noise describes a low-frequency brake noise from approximately 400 Hz to 500 Hz which arises at extremely low speeds and low brake pressures. Manoeuvres like slowly releasing the brake at a hill or gently braking against the drag torque of an automatic gearbox lead to honk noise. Under the same conditions, we observed creep groan at about 80 Hz. It has been shown that honk noise usually occurs after or alternates with creep groan. For this reason, it is assumed that honk noise—like creep groan—is a stick–slip-induced phenomenon and therefore shows highly nonlinear behaviour. In this paper, we present an approach for explaining the onset of honk noise under stick–slip excitation. A minimal model consisting of coupled mass oscillators excited by stick–slip is investigated. The model was able to reproduce the phenomena observed in the experiments. Thus, it is suitable for explaining the mechanisms leading to honk and estimate the influence of basic parameter variations. The lessons learned are a crucial step towards more realistic finite element or multi-body simulation methods, which have high potential for saving costs in the noise, vibration, and harshness (NVH) development process of brake systems
Characterisation of Recycled Concrete Aggregates with TGA and XRF coupled with Statistical Data Analysis
Get PDFA Review of Challenges and Opportunities in Occupant Modeling for Future Residential Energy Demand
Get PDFElectrified heating and mobility, the uptake of air conditioning and distributed energy resources are reshaping residential electricity demand and will require substantial investment. Yet the dependencies that drive present and future residential demand across sociodemographic characteristics, occupant activities, energy service demands, local technologies, and interactions with the overarching energy system remain poorly understood. Activity-based, bottom-up models make these dependencies explicit, better informing flexible operation and investment in low-carbon technologies.
We review 45 activity-based residential models and assess coverage of appliances, domestic hot water, space heating and cooling, and mobility (electric vehicle charging), which are rarely considered jointly in one integrated model. To our knowledge, this is the first review to include activity-based mobility modeling, thereby identifying methodological gaps in consistent behavior modeling across residential energy services: First, most studies simulate single occupants in isolation rather than entire households, thereby overlooking interdependencies among occupants. Second, predominant use of Markov models or independent univariate sampling limits temporal consistency. Based on these findings, future studies should combine complementary behavioral datasets with sophisticated models (e.g., deep neural networks) capable of capturing complex dependencies to generate high-quality synthetic behavioral data as a basis for future bottom-up residential energy demand modeling. Further progress requires open datasets and reproducible validation frameworks to benchmark and compare activity-based models and to ensure consistent progress in the field. Currently, there is no model available in the literature that derives energy demand for thermal comfort, hot water, mobility, and other services consistently from one fundamental representation of household behavior
FDO-Ops: A Model for Machine-Actionable FAIR Digital Objects to Enhance Interoperability
Get PDFInteroperability is the key to interacting with data across systems and technologies, thereby facilitating knowledge interpretation in different scientific domains. The FAIR Principles provide guidelines for this aspect and Knowledge Organization Systems and associated technologies provide solutions for this problem. However, interoperability is rarely fully achieved in reality due to limited capabilities with respect to machine actionability oftechnologies.
FAIR Digital Objects, when associated with operations, can fill this gap. We introduce FDO-Ops, a model that includes a specification of operations and a procedure for associating and executing them on FAIR Digital Objects. On this basis, we derive a generic description of the machine-actionable entity a FAIR Digital Object constitutes. To demonstrate our approach, we provide a prototype implementation of FDO-Ops that we test with two use cases from the domains of energy research and digital humanities. Our evaluation indicates that machine-actionable FAIR Digital Objects based on FDO-Ops, used in conjunction with existing technologies, greatly enhance the interoperability of the digital resources they represent at the technical and syntactic levels. In addition, they offer perspectives for improving semantic, legal, and organizational interoperability
Efficient Perovskite/Silicon Tandem Solar Cells Using Hybrid Two‐Step Inkjet Printing with Edge Isolation Precision
Get PDFDeveloping high-efficiency perovskite/silicon tandem solar cells (PSTs) using scalable deposition methods is crucial for the industrialization of next-generation photovoltaics. However, developing industrially viable deposition techniques to ensure high performance, uniformity, and compatibility with existing silicon manufacturing remains a key challenge. A scalable hybrid two-step deposition process, combining evaporation and inkjet printing, is presented for fabrication of high-performance PSTs. Wide bandgap perovskite solar cells are achieved with power conversion efficiencies (PCEs) of up to 19.8%. Applying this approach to textured silicon bottom cells, the process ensures conformal perovskite growth, critical for industry-relevant tandem integration. Using this technique, highly efficient, fully textured PSTs with a PCE of 27.4% are fabricated. Homogeneous perovskite thin films are formed up to the substrate\u27s very edge, enabling industry standards for silicon edge isolation. These results highlight the potential of hybrid two-step inkjet printing for scalable, high-efficiency PST fabrication, paving the way for industrial adoption
Dissecting Biomolecular Interactions Using Deep Learning Approaches
Get PDFImmunglobuline spielen sowohl bei biologischen Prozessen als auch bei
biotechnologischen Anwendungen eine wichtige Rolle, insbesondere aufgrund ihrer zentralen
Bedeutung für Immunreaktionen und die Entwicklung therapeutischer Arzneimittel. Um die
Immunologie voranzubringen und gezielte Therapien zu entwickeln, ist es entscheidend zu
verstehen, wie diese Proteine interagieren. Trotz bedeutender Fortschritte in der
Molekularbiologie bleibt die genaue Vorhersage und Beeinflussung von Protein-Protein-
Interaktionen (PPIs) eine Herausforderung, die den Durchbruch in Bereichen wie der
Arzneimittelentdeckung und Diagnostik behindert. Das Aufkommen fortschrittlicher Deep-
Learning-Techniken eröffnet jedoch neue Möglichkeiten zur Überwindung dieser
Herausforderungen, indem es schnelle und präzise Vorhersagen von PPIs ermöglicht. Diese
Studie ist motiviert durch das Potenzial dieser computergestützten Werkzeuge, unser
Verständnis von Immunglobulin-Interaktionen zu verbessern und dadurch zu Innovationen in
der Biotechnologie und therapeutischen Entwicklung beizutragen. In dieser Studie haben wir
ein weithin zugängliches neuronales Faltungsnetzwerk (Convolutional Neural Network, CNN)
angepasst, um eine speziesspezifische Klassifizierung verschiedener Immunglobulin G (IgG)
Komplexe zu erreichen. Wir haben Tröpfchen verschiedener Immunglobuline gemischt mit
dem B-Zell-Superantigen (SAg), rekombinantem Staphylokokkenprotein A, hergestellt und
auf hydrophobe Polymersubstrate aufgebracht. Diese Proteinfärbungen wurden dann mit Hilfe
der Polarisationslichtmikroskopie (PLM) abgebildet.
Unsere umfassende Analyse auf der Grundlage von 23.745 Bildern ergab, dass das
vortrainierte CNN, InceptionV3, nicht nur IgGs aus vier verschiedenen Spezies erfolgreich
kategorisierte, sondern auch ihre relative Bindungsstärke an Protein A vorhersagte. Im
Durchschnitt von 36 Bindungspaaren beobachteten wir (i) eine Gesamtgenauigkeit von 81.4%, (ii) die höchste Vorhersagegenauigkeit für humanes IgG, den Antikörper mit der höchsten
Bindungsaffinität für Protein A, und dass (iii) die Klassifizierungsgenauigkeit für die
verschiedenen IgG/Protein A-Verhältnisse im Allgemeinen mit der Bindungsstärke des
Protein-Protein-Komplexes korreliert, die mittels Circulardichroismus-Spektroskopie (CD)
bestimmt wurde. Darüber hinaus wurde das CNN, das ursprünglich mit IgG/Protein AFarbbildern
trainiert wurde, mit einem neuen Satz von Bildern getestet, bei denen ein anderes
Superantigen, rekombinantes Protein G, verwendet wurde. Bemerkenswerterweise
klassifizierte das CNN trotz des ungewohnten Superantigens die Bindungsstärke von humanem
IgG und Protein G korrekt und erreichte eine Genauigkeit von 94 % über verschiedene molare
Bindungsverhältnisse hinweg, da der ähnlichste IgG-Komplex im Trainingsdatensatz
vorhanden war.
Darüber hinaus wurde eine graphentheoretische Analyse eingesetzt, um den
bildbasierten Ansatz mit einer parameterbasierten neuronalen Netzwerkstrategie zu ergänzen.
Dieser innovative Ansatz wurde durch die Beobachtung von strukturellen Kristallmustern in
verschiedenen Proteinproben inspiriert. Die Graphentheorie, die für ihre vielseitigen
Anwendungen in verschiedenen wissenschaftlichen Disziplinen bekannt ist, wurde eingesetzt,
um Bilder in Graphen umzuwandeln. Mit Hilfe des an der Universität von Michigan
entwickelten Python-Pakets StructuralGT extrahierten wir aus diesen Graphen eine Reihe
aussagekräftiger Merkmale, die als Eingabedatensätze dienten. Diese Methode wurde mit den
Ergebnissen herkömmlicher neuronaler Faltungsnetzwerke verglichen.
Die Studie ergab, dass durch die Verwendung der aus den Graphen abgeleiteten
Merkmale als Eingabedatensatz für ein entwickeltes neuronales Netz die erforderliche
Trainingszeit im Vergleich zur bildbasierten Klassifizierung erheblich (etwa um das Dreifache)
reduziert werden konnte, wobei die Genauigkeit im optimierten Schema dennoch hoch blieb.
Die Ergebnisse unterstreichen das Potenzial der Kombination von Graphentheorie und Deep Learning für die Proteininteraktionsanalyse. Geeignete graphbasierte Merkmale können
zur Vorhersage von Protein-Protein-Interaktionen über den ursprünglichen Trainingsdatensatz
hinaus verwendet werden. Dieser Ansatz wird durch die Verarbeitung numerischer Daten
vereinfacht und ermöglicht die Klassifizierung auf nicht-GPU-abhängigen Systemen, was
sowohl die Rechenkosten als auch die Trainingszeit reduziert. Die Ergebnisse deuten auf eine
vielversprechende Methode zur Klassifizierung von biologischen Graphen und zur Vorhersage
der Stärke von Proteininteraktionen hin, die für das Protein-Engineering, das Verständnis der
Selbstaggregation und die Aufrechterhaltung der Proteinstabilität in komplexen Umgebungen
von Nutzen ist
A low-cost experimental averaging method for fast–slow mechanical systems via long-exposure video: Application to the Kapitza pendulum
Get PDFNonlinear mechanical systems are typically challenging to analyze experimentally, particularly when high-frequency motions require high temporal resolution, often making data collection difficult and expensive. This research explores an alternative approach by utilizing inexpensive and readily available long-exposure cameras, such as those found in billions of mobile devices. The method is suited to fast–slow systems in which a periodic, high-frequency motion rides on a much slower drift—examples include vibratory energy harvesters, tuned-mass dampers, and the Kapitza pendulum studied here. When the camera’s exposure covers at least one fast period, pixel brightness becomes proportional to the classical probability density (CPD) of the fast motion. Substituting this CPD into the standard averaging integral yields the governing equations of the slow subsystem without resolving the high-frequency motion in time. We demonstrate the experimental feasibility of the idea on a Kapitza pendulum whose pivot vibrates at
48 Hz. The blurred motion of three LEDs is captured on a 30fps HD video. For each frame, the spatial average is evaluated in MATLAB
, allowing for the recovery of the slow trajectory. Consequently, a system parameter identification is performed. Furthermore, with some uncertainty, it is even possible to recover the slow system dynamics of the unexcited, physical pendulum without directly measuring it.
Long-exposure imaging thus extracts relevant slow-dynamics information with pocket-sized hardware. It eliminates the need for costly high-speed cameras, providing a practical and low-budget tool for experimental averaging and system identification