67639 research outputs found
Sort by
Building simplified cancer subtyping and prediction models with glycan gene signatures
No full textWe identified a gene panel comprising 71 glycosyltransferases (GTs) that alter glycan patterns on cancer cells as they become more virulent. When these cancer-pattern GTs (CPGTs) were run through an algorithm trained on The Cancer Genome Atlas, they differentiated tumors from healthy tissue with 97% accuracy and clustered 27 cancers with 94% accuracy in external validation, revealing each variety’s “biometric glycan ID.” Using machine learning, we built four models for cancer classification, including two for detecting the molecular subtypes of breast cancer and glioma using even smaller CPGT sets. Our results reveal the power of using glyco-genes for diagnostics: Our breast cancer classifier was almost twice as effective in independent testing as the widely used prediction analysis of microarray 50 (PAM50) subtyping kit at differentiating between luminal A, luminal B, HER2-enriched, and basal-like breast cancers based on a comparable number of genes. Only four GT genes were needed to build a prognostic model for glioma survival.This work was supported by a King Abdullah University of Science and Technology (KAUST) Faculty Baseline Research Funding Program to J.S.M., by a KAUST Smart Health Initiative Grant (REI/1/4448-01) to J.S.M., by a KAUST AI Initiative Grant (REI/1/0018-01-01) to X.G., and by CBRC CCF (FCC/1/1976-23-01 and FCC/1/1976-26-01) to X.G. The research reported in this publication was also supported by funding from KAUST Center of Excellence for Smart Health (KCSH) under award number 5932. The results included here make use of data from TCGA Research Network (https://www.cancer.gov/tcga) and The Metastatic Breast Cancer Project (https://www.mbcproject.org/), a project of Count Me In (https://joincountmein.org/). We would also like to thank Daliah Merzaban for the professional writing/editing of the manuscript
Meta-Learning Based CTR Algorithm Selection and Hyperparameter Optimization
No full textThe existing Click-Through Rate (CTR) algorithms have their own advantages and are sensitive to hyperparameters. Quickly obtaining a high-performance CTR model for the new task can bring good application effects. However, ordinary users fail to do so due to the lack of domain knowledge. In this paper, we remedy this deficiency by proposing AutoCTR, an efficient meta-learning based Combined Algorithm Selection and Hyperparameter Optimization (CASH) algorithm, to help non-expert users quickly find the best CTR model. In AutoCTR, we introduce the meta-learning technique to make full use of the meta-information w.r.t. CTR to guide for the new CTR task. Specifically, we utilize the meta-information to learn characteristics and representations of CTR algorithms with different settings. We use these meta experiences combined with few evaluation information on the target CTR dataset to efficiently exploring the huge CTR CASH search space for the new task. The CTR model representation method has significant influence on the quality of the learned meta experiences. To further enhance the experiences quality, we also design a Graph Neural Network (GNN) based embedding learning method. This method can link different CTR models through their components, and thus quickly learning higher-quality model representations. Extensive experimental results show that AutoCTR can quickly select suitable CTR models for different CTR tasks. Compared with the existing CASH algorithms, which ignore meta-information or rely on a huge amount of meta-information, AutoCTR is more reasonable and efficient
Relations between Decision Trees and Decision Rule Systems
No full textThe study of relationships between systems of decision rules and deterministic decision trees is an important task of computer science. It is easy to transform a decision tree into a decision rule system. The inverse transformation is a more difficult task.
In this work, we study unimprovable upper and lower bounds on the minimum depth of decision trees derived from decision rule systems depending on the various parameters of these systems. To illustrate the process of transformation of decision rule systems into decision trees, we generalize a well-known result for Boolean functions to the case of functions of -valued logic.
We address six problems related to the inverse transformation. We demonstrate the complexity of constructing complete decision trees can be superpolynomial in many instances. We proposed three polynomial-time algorithms for each of the six problems, which focus on outlining the computation path within the decision tree for a specified input rather than constructing the entire tree. Additionally, we introduced a dynamic programming algorithm that calculates the minimum depth of a decision tree corresponding to a given decision rule system. We describe these algorithms and related to them theoretical results for each of the six problems and experimentally compare the performance of the three algorithms and evaluate their outcomes against the optimal results generated by the dynamic programming algorithm.
Finally, for conventional problems over an arbitrary infinite binary information system, we study relations between time and space complexity of deterministic and nondeterministic decision trees solving these problems and using only attributes from the problem descriptions. As time and space complexity, we consider the depth and the number of nodes in the decision trees. In the worst case, with the growth of the number of attributes in the problem description, (i) the minimum depth of deterministic decision trees grows either as a logarithm or linearly, (ii) the minimum depth of nondeterministic decision trees either is bounded from above by a constant or grows linearly, (iii) the minimum number of nodes in deterministic decision trees has either polynomial or exponential growth, and (iv) the minimum number of nodes in nondeterministic decision trees has either polynomial or exponential growth. Based on these results, we divide the set of all infinite binary information systems into three complexity classes and study for each class issues related to time-space trade-off for decision trees
Efficient Glycerol Ketalization with Acetone Catalyzed by Nanosized Fe-Doped Aluminoborate PKU-1
No full textGlycerol ketalization has gained growing interest for glycerol recycling under mild reactions. To improve the product yield, high-purity glycerol and excessive acetone have usually been used, resulting in high production costs. Here, Fe-doped aluminum borate (Fe-PKU-1) materials with tunable surface physicochemical properties were synthesized by varying aluminum reactants and adjusting the loading amount of cetyltrimethylammonium bromide (CTAB). The hydrophobicity–hydrophilicity balance and surface acidity of these materials both display volcano trends with variation over material morphology and CTAB loading amounts. As a result, 40 mg of CTAB-modified nanosized Fe-PKU-1 achieved a higher catalytic efficiency for glycerol ketalization, i.e., a 97.2% conversion efficiency of pure glycerol and a 98.0% selectivity of (2,2-dimethyl-1,3-dioxolan-4-yl) methanol, with less acetone doping (acetone/glycerol = 3.5:1) in a 3 h reaction time. Moreover, the synergy between improved surface hydrophobicity and material acidity suppresses the influence of CH3OH and NaCl impurities in crude glycerol on the reaction, resulting in an 84.7% conversion efficiency of crude glycerol to the target compound. This low-cost catalyst system demonstrates good potential for the application in the glycerol industry.This work was financially supported the Doctoral Fund of Ministry of Education of China (grant no. 2023M730428), Chongqing Postdoctoral Foundation (grant no. 2021XM3006), National Natural Science Foundation of China (grant no. 52220105010), and Guangdong Basic and Applied Basic Research Foundation (grant no. 2023A1515010052)
Elucidating the photodegradation pathways of polymer donors for organic solar cells with seven months of outdoor operational stability
No full textThe operating lifetime under real-world climates is a critical metric to evaluate the commercial potential of any photovoltaic technology. Organic solar cells (OSCs) have experienced rapid breakthroughs in performance over the past decade owing to advances in device and materials engineering, including interfaces, electron acceptors, and donors. However, the intrinsic photodegradation of polymer donors remains poorly understood, and a path to stable OSCs is yet to be demonstrated under outdoor testing conditions. Herein we elucidate the side-chain-induced degradation mechanism in polymer donors and present an outdoor stability database covering 15 representative non-fullerene-based OSCs, supported by in-lab photostability and thermostability analysis. By understanding the performance losses induced by several photoactive layers and interfaces, we demonstrate that encapsulated non-fullerene-based OSCs can retain 91% of the initial efficiency after seven months of operation under hot and sunny Saudi Arabian climates. These findings reveal encouraging prospects of non-fullerene-based OSCs for outdoor applications.This publication is based on work supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under award no. CCF-3079. J.H. expresses gratitude to the Alexander von Humboldt Foundation and the support during his stay in T. B. Marder’s group at Julius-Maximilians-Universität Würzburg. H.X. would like to extend thanks to A. V. Marsh and M. Heeney for providing the training and technical support related to the GPC instrumentation. We would like to thank the KAUST weather team for providing access to weather station data. We acknowledge the use of the KAUST Solar Center and the support from its staff
New Monolithic Modular DC-DC Converter for Constant Current Offshore Interconnection with VSC HVDC Network
No full textThe rapid growth in HVDC (High Voltage Direct Current) grids has shown the falls of point-to-point connections. Several challenges such as the requirement of DC (Direct Current) fault blocking capability, interfacing of different grounding schemes, offering multi-vendor interoperability, and difficult to achieve high DC voltage stepping, represent serious issues to deployment of HVDC grids. DC-DC converters are considered the optimum candidate to overcome these challenges in HVDC grids interconnection. In this paper, a novel isolated hybrid monolithic modular DC-DC converter is proposed that interconnects LCC/VSC (Line Commutated Converter/Voltage Source Converter) based HVDC networks. It achieves smaller count of semiconductors, lower conduction losses and DC fault blocking. Detailed mathematical analysis, design, and control of the proposed DC-DC converter are illustrated. Also, both simulation model and experimental test rig are built to validate the proposed DC-DC converter under different normal operational and fault scenarios
Understanding the role of autoencoders for stiff dynamical systems using information theory
No full textUsing information theory, this study provides insights into how the construction of latent space of autoencoder (AE) using deep neural network (DNN) training finds a smooth (non-stiff) low-dimensional manifold in the stiff dynamical system. Our recent study (Vijayarangan et al. 2023) reported that an AE combined with neural ODE (NODE) as a surrogate reduced order model (ROM) for the integration of stiff chemically reacting systems led to a significant reduction in the temporal stiffness, and the behavior was attributed to the identification of a slow invariant manifold by the nonlinear projection using the AE. The present work offers a fundamental understanding of the mechanism of formation of a non-stiff latent space and stiffness reduction by employing concepts from information theory and better mixing. The learning mechanisms of both the encoder and the decoder are explained by plotting the evolution of mutual information and identifying two different phases. Subsequently, the density distribution is plotted for the physical and latent variables, which shows the transformation of the rare event in the physical space to a highly likely (more probable) event in the latent space provided by the nonlinear autoencoder. Finally, the nonlinear transformation leading to density redistribution is explained using concepts from information theory and probability.This work was funded by King Abdullah University of Science and Technology (KAUST) and utilized the computational resources of the KAUST Supercomputing Laboratory (KSL)
Multimethod analysis of habitat connectivity and foraging in seagrass-based megafauna
No full textSeagrass meadows are foundational coastal ecosystems that sustain biodiversity, support fisheries, and store blue carbon, yet their role in green turtle diets in the Red Sea and their contribution to tiger shark trophic ecology in The Bahamas remain unquantified. This thesis investigates how seagrass ecosystems support mobile herbivores and apex predators, and the implications for their conservation across regions and jurisdictions. Focusing on green turtles (Chelonia mydas) in the Red Sea and tiger sharks (Galeocerdo cuvier) in The Bahamas, the research integrates spatial ecology, stable isotope analysis, and governance assessments to provide a multimethod analysis of habitat connectivity and trophic interactions.
The first component examined the spatial ecology of Red Sea green turtles, revealing both local foraging migrations and transboundary movements across five Exclusive Economic Zones (EEZs). While most individuals foraged within Saudi Arabia’s EEZ, turtles were also recorded in Egypt, Sudan, Eritrea, and Yemen, showing that populations are shared resources. Network centrality metrics identified Sudan as an important linking node. Standardising distance effects in multimodal models revealed a link between foraging probability and EEZ coastline length, while a conservation enforcement capacity index highlighted major disparities among states.
The second component assessed turtle diets, showing strong reliance on seagrass species such as Cymodocea rotundata and Enhalus acoroides. Stable isotope analysis suggested potential overgrazing in some meadows, raising concerns for ecosystem functioning and carbon storage. Green turtle isotopic data also show promise as an indirect monitoring tool for seagrass status.
The third component analysed tiger shark foraging, demonstrating that seagrass-derived carbon contributed substantially (~23%) to their longer-term diet. Ontogenetic shifts were evident, with smaller sharks relying on nearshore seagrass and reef resources, while larger individuals targeted pelagic prey. These results illustrate tiger sharks as ecological connectors, linking seagrass with offshore food webs.
Together, the findings highlight seagrass ecosystems as critical nodes sustaining multiple trophic levels. By integrating ecological, spatial, and governance perspectives, the thesis underscores the need for cross-jurisdictional collaboration, capacity building, and habitat protection to support biodiversity, livelihoods, and global climate and conservation targets, including the Kunming–Montreal Global Biodiversity Framework
Charting the single cell transcriptional landscape governing visual imprinting.
No full textMemory-related transcriptional events in brain remain poorly understood. Visual imprinting is a form of learning in which young animals develop preferences through early exposure to specific stimuli. In chicks, visual imprinting memory is stored in the intermediate medial mesopallium (IMM) of the forebrain. To investigate learning-associated molecular changes, we performed single-nucleus RNA sequencing of the left IMM in strongly imprinted chicks and untrained controls. This analysis identified over 30 cell clusters with distinct transcriptional differences putatively linked to memory formation, nearly half of them in long non-coding RNAs (lncRNAs). Expression levels of two lncRNAs and four protein-coding genes FOXP2, RORA, LUC7L, and ROBO1 correlate with memory strength, reflecting either innate learning potential or imprinting experience. Notably, the brain- and avian-specific lncRNA ENSGALG00010007489 is enriched in the nuclei of specific glutamatergic clusters. These findings offer the first single-cell resolution map of transcriptional changes underlying memory formation in the avian brain.This work has been funded by (i) the European Union's Horizon 2020 research and innovation program (project CHARM-Vis, project ID 867429), (ii) by the Shota Rustaveli National Science Foundation, project number NFR-22-8692, and (iii) by basic funding from Ilia State University and King Abdullah University of Science and Technology
Mechanism of cooperative strigolactone perception by the MAX2 ubiquitin ligase–receptor–substrate complex
No full textStrigolactones are plant hormones that regulate development and mediate interactions with soil organisms, including the germination of parasitic plants such as Striga hermonthica. Strigolactone perception by receptors initiates the degradation of transcriptional repressors via E3 ubiquitin ligases, but the mechanistic link between hormone binding and substrate ubiquitination has remained unclear. We determine cryogenic electron microscopy structures of the receptor–ligase–substrate complex, composed of Arabidopsis ASK1 and substrate, and Striga F-box and receptor proteins. Strigolactone hydrolysis by the receptor, which covalently retains the D-ring, is a prerequisite for complex formation. The substrate engages the complex through two domains, forming a dynamic interface that stabilises the receptor–ligase assembly and repositions the ASK1, suggesting a mechanism for efficient ubiquitination. Here, we show how dynamic, multivalent interactions within the receptor–ligase–substrate complex translate hormone perception into targeted protein degradation, providing insight into how plants integrate hormonal signals into developmental decisions.We thank L. Zhao for help with cryo-EM data recording. We thank S. Al-Babili and A. De Biasio for valuable discussions and S. Al-Babili for reagents. For computer time, this research used the resources of the KAUST Supercomputing Laboratory, and experimental research was supported by the Bioscience Core Lab, ACL Proteomics Lab and the Imaging and Characterisation Core Lab at King Abdullah University of Science & Technology (KAUST) in Thuwal, Saudi Arabia. This research was supported by the King Abdullah University of Science and Technology (KAUST) through the baseline fund to STA and the Award No. URF/1/4039-01-01 and URF/1/4080-01-01. We acknowledge the support by the German Research Council (DFG) through the core facility for HDX-MS (projects 260989694 and 324652314 to Gert Bange, Philipps-University Marburg)