Ghent University Global Campus

Sub-visible particle analysis using flow imaging microscopy combined with deep learning has proven effective in identifying particle types, enabling the distinction of harmless components such as silicone oil from protein particles. However, the scarcity of available data and severe imbalance between particle types within datasets remain substantial hurdles when applying multi-class classifiers to such problems, often forcing researchers to rely on less effective methods. The aforementioned issue is particularly challenging for particle types that appear unintentionally and in lower numbers, such as silicone oil and air bubbles, as opposed to protein particles, where obtaining large numbers of images through controlled settings is comparatively straightforward. In this work, we develop a state-of-the-art diffusion model to address data imbalance by generating high-fidelity images that can augment training datasets, enabling the effective training of multi-class deep neural networks. We validate this approach by demonstrating that the generated samples closely resemble real particle images in terms of visual quality and structure. To assess the effectiveness of using diffusion-generated images in training datasets, we conduct large-scale experiments on a validation dataset comprising 500,000 protein particle images and demonstrate that this approach improves classification performance with no observable downside. Finally, to promote open research and reproducibility, we publicly release both our diffusion models and the trained multi-class deep neural network classifiers, along with a straightforward interface for easy integration into future studies, at https://github.com/utkuozbulak/svp-generative-ai.

Background/Objectives: Translating brain volumetric biomarkers to individual-level Alzheimer's disease (AD) diagnosis remains challenging due to difficulty interpreting raw volumes without longitudinal monitoring or matched controls. We tested a classification model using population-referenced volumetric percentiles to distinguish AD from cognitively normal (CN) subjects and evaluated its generalization across independent cohorts. Methods: Brain volumes from 95 regions were extracted using an automated segmentation pipeline and converted to age and sex adjusted percentiles using a reference population (N = 1833). A logistic regression classifier was trained on ADNI subjects (N = 873; AD = 183, CN = 690) split into training (60%), validation (20%), and test (20%) sets. The model was evaluated on two independent validation datasets: the held-out ADNI validation set and an external Korean cohort (N = 72; AD = 36, CN = 36) acquired with different scanner protocols and demographic characteristics. Results: The model achieved excellent discrimination across all evaluation sets: ADNI validation (AUC = 0.963, accuracy = 90.3%), ADNI test (AUC = 0.960, accuracy = 89.7%), and Korean external validation (AUC = 0.981, accuracy = 87.5%). The minimal validation gap (0.018) demonstrated robust generalization. Positive coefficients for ventricular regions reflected AD-associated atrophy patterns, while negative coefficients for medial temporal structures indicated their contribution within multivariate patterns distinguishing AD from normal aging. Conclusions: Population-referenced brain volumetric percentiles enable accurate AD classification with robust generalization across populations and scanner protocols. By contextualizing individual brain structure relative to normative populations while accounting for age and sex, this approach demonstrates potential for clinical translation as an accessible neuroimaging-based diagnostic tool.

This paper presents a comprehensive bibliometric analysis of scientific output in the field of Multi-Criteria Decision-Making (MCDM) within the context of computer science, focusing on the period from 2019 to mid-2025. Using data from the Web of Science Core Collection, a total of 302 relevant papers were identified based on criteria such as publication year, language, document type, open access status, and research area. The analysis covers publication dynamics, the most influential authors and institutions, thematic directions, and collaboration structures. Special emphasis is placed on citation analysis of the most impactful works, as well as the visualization of co-authorship networks using the PRISMA methodology and tools such as RStudio and Biblioshiny. The results reveal a growing trend in publications, high activity by certain authors (e.g., Akram M and Liu Y), and strong collaboration within research clusters. Dominant topics include decision-making models, alternative selection, aggregation operators, and priority evaluation. This study provides insights into the structure and dynamics of the scientific community engaged in MCDM methods in computer science and may serve as a guide for future researchers, practitioners, and scientific development strategies.

ABSTRACT Pre-mRNA splicing is carried out by the spliceosome, a large and dynamic ribonucleoprotein complex. The spliceosome is known to be stored in nuclear speckles (NS), which are now recognized as active subnuclear organelles for splicing. However, it remains poorly understood how spliceosomal protein-protein interactions are functionally coupled to NS organization to maintain splicing robustness in plants. Here, we report the functional significance of a specific interaction between two U4/U6·U5 tri-snRNP components of the spliceosome, STA1 and DOT2, in regulating NS organization, pre-mRNA splicing, and heat stress responses in Arabidopsis . We identified a missense mutation in DOT2 (a Snu66/SART1 homolog) from a genetic suppressor of the PRP6 homolog mutant sta1-1 (named S307 ). This mutation restored the weakened interaction between STA1 and DOT2 in the sta1-1 mutant background. Genetic, biochemical, and cell biological analyses showed that variation in the strength of the STA1-DOT2 interaction was closely associated with changes in NS formation, splicing efficiency, as well as growth and heat tolerance. Pharmacological inhibition of STA1-associated NS formation by tubercidin recapitulated sta1-1 -like phenotypes and splicing defects, supporting a functional link between NS organization and splicing outcomes. In addition, heat-induced weakening of the STA1-DOT2 interaction was accompanied by reduced NS formation and increased intron retention at the transcriptome-wide level including key heat-responsive transcripts. Based on these observations, we propose that the STA1-DOT2 interaction, likely reflecting the assembly state of the U4/U6·U5 tri-snRNP, functions as a heat-sensitive interaction node that couples spliceosome assembly to NS organization and splicing robustness under stress conditions.

HighlightsWhat are the main findings?Four f-TNA structural configurations (closed/open, up/down) were compared to optimize DSSC performance.The open f-TNA structure, with the barrier layer removed, showed the highest power conversion efficiency (PCE ≈ 7.73%).The best PCE (7.73%) was achieved using the open-up f-TNA photoanode configuration. Barrier layer removal in f-TNA effectively minimized charge transfer resistances (Rct1 and Rct2).The closed-up configuration yielded the lowest PCE (5.52%) due to poor electrolyte diffusion kinetics.What are the implications of the main findings?The structural configuration and orientation of f-TNA are critical design parameters for high-performance DSSC.Structural optimization alone is a highly effective, additive-free strategy for maxim-izing TiO2 nanotube DSSC efficiency.Opening the f-TNA bottom is essential for enhancing charge collection and improv-ing electrolyte penetration.The study presents fundamental guidelines for engineering f-TNA electrodes in next-generation photoelectrochemical devices.Dye-sensitized solar cells (DSSCs) are known for their excellent low-light performance, cost-effectiveness, and flexibility. The photoanode has a crucial role in enhancing the overall performance of DSSCs and can be modified with different nanostructures. This study explores the impact of photoanode structure on the power conversion efficiency (PCE) of DSSCs, where four configurations of freestanding TiO2 nanotube arrays (f-TNAs), closed-up, closed-down, open-up, and open-down, were employed as photoanodes. Performance was evaluated based on current density (Jsc), open-circuit voltage (Voc), fill factor (FF), and PCE concerning dye adsorption, electrolyte diffusion, electron transport, and barrier layer. DSSCs based on open configurations, open-up and open-down f-TNAs, demonstrated superior performance, achieving PCE of 7.73% and 7.71%, respectively. The primary distinction between the DSSCs based on open-up f-TNAs and those based on open-down f-TNAs lies in the dye adsorption time and electron diffusion characteristics. The PCE for DSSCs with closed-down f-TNAs was measured at 6.78%, while DSSCs with closed-up f-TNAs showed a lower PCE of 5.52%. The presence of a barrier layer under the bottom of f-TNAs impacted the PCE for DSSCs with closed-down f-TNAs, whereas for DSSCs with closed-up f-TNAs, insufficient dye loading, poor electrolyte diffusion and barrier layer reduced the performance.

Summary Iron is essential for the growth and metabolism of marine algae. It also participates in the biosynthesis pathway of saxitoxins in dinoflagellates. High iron concentration may enhance the biosynthesis of saxitoxin in Alexandrium species. To better understand the effect of iron on the growth and metabolism of Alexandrium species, the dominant red tide species in Korea, A. catenella , was cultivated at four different iron concentrations (1, 10, 100, and 1000 nM). The growth data show that, within the range from 1 to 100 nM, growth was positively correlated with the iron concentration. High iron concentration (1000 nM) inhibited the growth of A. catenella. The data from the photosynthesis analysis showed that iron enhanced the maximum relative electron transport rate when the concentration was not higher than 100 nM. A similar result was also found in chlorophyll a content. At the same time, the expressions of several genes related to photosynthesis and saxitoxin synthesis were analyzed at three iron concentrations (1, 10 and 100 nM). The results show that psbE , a gene encoding iron‐binding protein in photosystem II, was significantly upregulated at the iron concentration of 100 nM. Our results increase the knowledge of the effects of iron on the growth, photosynthesis, and gene expression of A. catenella .