324139 research outputs found
Sort by
Learning with higher-order interactions: from hypergraphs to multi-agent systems
Get PDFHigher-order interactions are defined by the capacity of multiple entities to collectively produce a meaningful group-level outcome, distinguishing their collaboration from a mere random assembly. They involve two key elements: the entity-level features of the entities involved and the group-level outcomes that emerge. Such interactions are pervasive across domains, from researchers co-authoring publications to programmers developing software. Therefore, it is crucial to explore learning methods that capture both sides of higher-order interactions: their interplay with entity-level features and their influence on group-level outcomes. This thesis addresses this challenge with a portfolio of novel methods at both the entity and group levels. At the entity level, we leverage the lens of hypergraph machine learning. We begin with the task of inferring implicit higher-order interactions from observed entity features, proposing HGSL, a hypergraph structure learning framework that infers implicit higher-order interactions with a novel dual-smoothness prior. For explicitly defined higher-order interactions, the key entity-level challenge is exploiting such interactions to enrich entity features. To this end, we propose two novel approaches: Hypergraph-MLP, which embeds higher-order interaction directly with a multi-layer perceptron via an original loss function, and TF-HNN, which uses a novel training-free message-passing module to encode higher-order interaction into entity features. Moving beyond the entity level, we turn to the study of higher-order interactions at the group level. We study this through the lens of large language model (LLM) based multi-agent systems. We first propose MATRIX, an LLM multi-agent simulation framework with a novel homophily-guided communication mechanism, showing that the group-level outcomes of higher-order interactions can lead to high-quality data for LLM posttraining. We then present a novel analysis framework revealing how the intrinsic characteristics of a task, especially its logical depth and capability width, critically determine whether higher-order interactions within LLM-based agents will yield effective group-level outcomes. To sum up, this thesis contributes to learning with higher-order interactions at both the entity and group levels from hypergraphs to multi-agent systems
Spatiotemporal tidal prediction and analysis through physics-informed machine learning: Spatiotemporal tidal prediction and analysis through physics-informed ML
Get PDFTidal processes—those driven by gravitational forcing from the Earth, Moon, and Sun—shape coastal environments and significantly impact infrastructure. While conventional methods of tidal analysis and prediction perform well for barotropic tides sampled at high temporal resolution, major challenges arise when exogenous non-tidal forcing is present or temporal sampling becomes sparse and irregular. Under these conditions, the accuracy of empirical methods degrades as their underlying assumptions are violated.This thesis introduces four novel methodologies that advance the prediction and analysis of tidal and tidally driven processes across diverse spatiotemporal scales. Central to each is the augmentation of classical empirical techniques with modern machine learning, reducing reliance on inaccurate assumptions and improving both predictive power and physical interpretability.The work focuses on two classical approaches to tidal analysis: harmonic analysis and the response method. Each has strengths that suit particular challenges. For harmonic analysis, I develop a Bayesian framework tailored to sparse, noisy satellite altimetry data. This approach enables the exploitation of Surface Water and Ocean Topography (SWOT) mission data, provides uncertainty estimates lacking in existing models, and led to the discovery of an earth-shaking wave using this dataset.For the response method, I develop an automated, non-parametric procedure—overcoming a major limitation of the classical approach. It is the only method capable of analyzing and predicting arbitrary tidal processes, including tidal rivers and storm surges. I further propose a coupled response theory for predicting tidal currents. This method offers superior predictive accuracy to harmonic analysis and reduces the required time-series duration by nearly an order of magnitude -- a feature that can enhance harmonic analysis as well.Finally, I consider augmenting all classical methods -- including harmonic, response, and numerical models -- with short-term forecasting using recent data via an online training procedure. These contributions address long-standing limitations in empirical tidal analysis and have broad applications -- from improving satellite data usage to enhancing operational storm surge forecasting. This thesis argues that the `tide problem' remains unsolved and offers tools and solutions to some of its most pressing challenges
Dataset to support article: Better material properties and faster catalysed chemical recycling for poly(L-lactide) using a simple commercial glycerol ethoxylate additive
No full textData to support article. Includes NMR spectra, TGA depolymerization profiles, thermomechanical and spectral characterisation of blends, starting materials and commercial polymers
Characterising nanobody developability to improve therapeutic design using the Therapeutic Nanobody Profiler
Get PDFDevelopability optimisation is an important step for successful biotherapeutic design. For monoclonal antibodies, developability is relatively well characterised. However, progress for novel biotherapeutics such as nanobodies is more limited. Differences in structural features between antibodies and nanobodies render current antibody computational methods unsuitable for direct application to nanobodies. Following the principles of the Therapeutic Antibody Profiler (TAP), we have built the Therapeutic Nanobody Profiler (TNP), an open-source computational tool for characterising nanobody developability. Tailored specifically for nanobodies, it accounts for their unique properties compared to conventional antibodies for more efficient development of this novel therapeutic format. We calibrate TNP metrics using the 36 currently available sequences from clinical-stage nanobody-based drugs. We also collected experimental developability data for 108 nanobodies expressed as IgG constructs and examine how these results are related to the TNP guidelines. TNP is available as a web application at opig.stats.ox.ac.uk/webapps/tnp
The global extent of the grassland biome and implications for the terrestrial carbon sink
Get PDFLand cover data are commonly used to model the terrestrial carbon (C) sink, yet these data have wide margins of error that significantly alter estimates of global C storage. Here we demonstrate this data vulnerability in grasslands, which are critical to C cycling but whose estimated distribution has varied by >50 million km2 (3.5–42% of the Earth’s terrestrial surface). Comparing multiple high-resolution land cover products with expertly annotated grassland data from six continents, we show sources of mapping error and discuss C implications based on 2023 United Nations (UN) FAO estimates. Past misidentification arose from inconsistent definitions on grassland identity and classification flaws especially relating to woody plant cover. Correcting these errors adjusted grassland coverage to 22.8% of the terrestrial land base (30.1 million km2), elevating UN projections of soil C stocks to 155.02 Pg (0–30 cm depth). These findings underscore the challenges of biome mapping for ecosystem accounting and policy, when lacking field-validated remotely sensed data
Assessing Aβ‐independent effects of Module 42 on immune function in vitro
Get PDFINTRODUCTION: A deep multi‐omic analysis of post mortem human brains has identified a new co‐expression protein network – Module 42 (M42), strongly corelated with Alzheimer's disease (AD) pathology. M42 comprises 32 transmembrane and extracellular matrix (ECM)‐associated proteins, including the amyloid precursor protein (APP) and apolipoprotein E (apoE), and its members have been implicated in amyloid beta (Aβ) pathology. We systematically evaluated the Aβ‐independent effects of M42 on immune function in vitro. METHODS: Recombinant M42 proteins were expressed and purified. Their effects on phagocytosis, intracellular signaling, and cell viability were assessed in human induced pluripotent stem cell‐derived macrophages. RESULTS: Treatment with Midkine (MDK) reduced phagocytosis, while treatment with the ectodomain of Transmembrane protein with EGF‐like and two follistatin‐like domains 2 (TMEFF2) had the opposite effect. Both proteins promoted intracellular Ca2+ signaling, and TMEFF2 also suppressed Syk kinase activity. No M42 proteins had an effect on viability. DISCUSSION: Our results suggest an additional role for M42 in AD via regulating immune functions. Highlights: We tested M42 proteins for their effects on immune functions in vitro. Five proteins altered phagocytosis, and seven altered Ca2+ signaling. MDK and TMEFF2 ectodomain had an effect on both phagocytosis and Ca2+ signaling