DeustoTeka
DeustoTeka recoge la producción científica del personal docente e investigador de la Universidad de Deusto. Su propósito es reunir, archivar, preservar y aumentar la visibilidad en acceso abierto de los resultados de investigación.
Comunidades en DSpace
Selecciona una comunidad para explorar sus colecciones.
Envíos recientes
Time-varying systemic risk in electricity markets using generative adversarial networks: market resilience and policy
(Elsevier Ltd, 2026-03) Bohórquez Correa, Santiago
; Mosquera López, Stephanía
; Uribe Gil, Jorge Mario
Current frameworks and policy instruments for monitoring and regulating European electricity markets fall short of fully addressing the complexities that arise during periods of market distress. Our study makes two key contributions in this area: First, it provides a novel, integrative analysis of systemic risk across 25 energy markets, encompassing oil, natural gas, and coal, as well as 21 European electricity markets. Second, it introduces Time Series Generative Adversarial Networks to systemic risk literature, enabling real-time tracking of market resilience. Our findings show that systemic distress in European electricity markets was higher in Q3 2021 than in late 2021 and early 2022, despite record-high electricity prices in the latter period, which many assumed reflected maximum market distress. This suggests that policy interventions enacted at the end of 2021 effectively reduced systemic distress in European electricity markets. However, fossil fuel markets reached a peak in risk during the first quarter of 2022, underscoring energy security concerns for Europe due to its reliance on foreign fuel sources, as prices are set in a global rather than regional context. Our modeling framework offers a tool to assess such risks in real time, providing valuable insights for proactive policymaking in the European energy sector.
Identification of mathematical patterns in genomic spectrograms linked to variant classification in complete SARS-CoV-2 sequences
(Nature Research, 2025-12-05) Guerrero Tamayo, Ana; Sanz Urquijo, Borja; Moragues Tosantos, María Dolores; Olivares, Isabel; Casado, Concepción; Pastor López, Iker
Building on previous studies, we identified mathematical patterns in HIV-1 and SARS-CoV-2 genomes using transfer learning and explainability with a pre-trained CNN on genomic spectrograms. These patterns seemed to define viral characteristics, leading us to hypothesize that inherent mathematical patterns in a virus’s genome determine its features. To explore this further, we focused on SARS-CoV-2 variant classification, designing a methodology with genomic spectrograms, a two-stage transfer learning approach, and two-step explainability. This approach identified genomic regions and nucleotide frequency patterns that characterize specific variants, revealing clear, distinguishable patterns for each category. The distinct and consistent total regions of high activation for each variant highlight the significance of the genomic region from the beginning of S gene to the end of 3’UTR in identifying the variants under study. The frequencies and particularly within this region appeared to play a key role in their identification. The shared prominence of in the final segment of the genome for both pre-VOC and Omicron (despite different pattern shapes) may hint at a phylogenetic connection in SARS-CoV-2 or even suggest that Omicron evolved from a pre-VOC lineage. The confirmation that mathematical patterns are associated with variant classification represents a step forward in demonstrating that these patterns play a role in viral characterization, suggesting the existence of an additional layer of genomic information that may enable virus characterization in a low-computing, and efficient manner compared to traditional methodologies.
Neural damage and inflammation in myotonic dystrophy type 1: longitudinal analysis of serum NFL, GFAP, and IL-6
(Elsevier Inc., 2026-01) Garmendia, Joana; Labayru, Garazi
; Alberro, Ainhoa; Martins-Almeida, Laura; Otaegui Bichot, David; Iruzubieta Agudo, Pablo
; López de Munain Arregui, Adolfo; Sistiaga Berrondo, Andone
Introduction: Myotonic dystrophy type 1 (DM1) is a progressive, multisystemic disease affecting the central nervous system (CNS). Blood-based biomarkers such as neurofilament light chain (NFL), glial fibrillary acidic protein (GFAP), and interleukin-6 (IL-6) offer potential as non-invasive indicators of CNS dysfunction and/or inflammation. However, their longitudinal dynamics and clinical relevance in DM1 remain unclear. Additionally, sex-related differences in these biomarkers are poorly understood. This study aimed to investigate NFL, GFAP, and IL-6 serum levels in patients with DM1, examine sex-differences, track changes over four years, and explore associations with genetic, muscular, cognitive, and neuroimaging outcomes. Method: Retrospective data from 70 DM1 patients and 54 healthy controls (HC) were analyzed. Longitudinal data were available for 68 participants (39 DM1, 29 HC). Biomarkers were measured using the ELLA immunoassay. DM1 patients had data on genetic, muscular, cognitive and structural brain outcomes. Analyses were adjusted for age. Results: NFL and IL-6 levels were significantly higher in DM1 patients compared to HC, while GFAP levels did not differ. Male DM1 patients exhibited higher NFL and IL-6 levels compared to females. No significant longitudinal changes were observed over a four-year period. NFL and IL-6 levels correlated with larger genetic expansions and poorer cognitive performance. Discussion: NFL and IL-6 may reflect neural damage and systemic inflammation in DM1 and could serve as biomarkers of cognitive dysfunction. However, their limited longitudinal sensitivity suggests longer follow-up is needed to evaluate their utility for disease monitoring.
D-CRISP: explaining object detectors by combining randomized and segment-based perturbations
(IOS Press BV, 2025-10-21) Andrés Fernández, Alain; Ser Lorente, Javier del
Explaining the decisions issued by Machine Learning models for object detection tasks is essential in high-stakes decision making scenarios, such as medical image processing and vehicular perception for autonomous driving. Despite the proliferation of post-hoc perturbation-based methods for generating visual explanations, most eXplainable AI (XAI) approaches rely exclusively on either random image masking or selective segmentation-based occlusion, missing the opportunity to synergistically leverage both strategies in a complementary fashion. In this paper we address this gap by proposing D-CRISP (Detector-Combining Randomized Input and Segment Perturbations), a novel post-hoc explanation method for object detection models. D-CRISP unifies both random and region-based occlusions derived from image segmentation, producing multiscale saliency maps that capture both granular (pixel-level) and semantic (region-level) cues about the objects detected by the model. Experiments on the MS-COCO dataset show that D-CRISP significantly outperforms random-masking approaches in terms of explanation faithfulness and localization, while requiring slightly more computation effort than these methods. At the same time, it achieves comparable or better performance than segmentation-based methods, yet with substantially lower mask generation latencies. These results position D-CRISP as a highly effective and efficient XAI alternative for object detection models, particularly suited for time-constrained applications requiring timely, accurate, and interpretable decisions.
Influence of the sample shape factor on the dynamic characterization of viscoelastic properties: complex moduli and Poisson's ratio
(Elsevier Ltd, 2026) Cortazar Noguerol, Julen
; Cortés Martínez, Fernando
; Elejabarrieta Olabarri, María Jesús
This study investigates how the sample shape factor influences the dynamic properties characterization of a silicone rubber within the linear viscoelastic regime. The effective elastic properties of elastomers are known to depend on geometry, but the effect of shape factor on the dynamic response has not been systematically characterized. To address this, cylindrical samples with varying geometries are tested under dynamic compression and torsion. The results reveal that both the complex compressive and shear moduli are affected by shape factor, and that this influence varies with frequency. To quantify the influence of shape factor and extract the material's dynamic properties, a phenomenological correction model is formulated. The model introduces frequency-dependent parameters that account for the geometric effects on the effective moduli. These corrected moduli yield a complex Poisson's ratio that exhibits a slight frequency dependence, with a decreasing real part and an increasing loss factor. This approach enables both the quantification of geometry-induced effects in dynamic mechanical testing and the extraction of intrinsic material's viscoelastic properties.