Browse Articles
Discover research articles across all indexed journals
A multimodal deep learning architecture for predicting interstitial glucose for effective type 2 diabetes management
Abstract The accurate prediction of blood glucose is critical for the effective management of diabetes. Modern continuous glucose monitoring (CGM) technology enables real-time acquisition of interstitial glucose concentrations, which can be calibrated against blood glucose measurements. However, a key challenge in the effective management of type 2 diabetes lies in forecasting critical events driven by glucose variability. While recent advances in deep learning enable modeling of temporal patterns in glucose fluctuations, most of the existing methods rely on unimodal inputs and fail to account for individual physiological differences that influence interstitial glucose dynamics. These limitations highlight the need for multimodal approaches that integrate additional personalized physiological information. One of the primary reasons for multimodal approaches not being widely studied in this field is the bottleneck associated with the availability of subjects’ health records. In this paper, we propose a multimodal approach trained on sequences of CGM values and enriched with physiological context derived from health records of 40 individuals with type 2 diabetes. The CGM time series were processed using a stacked Convolutional Neural Network (CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) network followed by an attention mechanism. The BiLSTM learned long-term temporal dependencies, while the CNN captured local sequential features. Physiological heterogeneity was incorporated through a separate pipeline of neural networks that processed baseline health records and was later fused with the CGM modeling stream. To validate our model, we utilized CGM values of 30 min sampled with a moving window of 5 min to predict the CGM values with a prediction horizon of (a) 15 min, (b) 30 min, and (c) 60 min. We achieved the multimodal architecture prediction results with Mean Absolute Point Error (MAPE) between 14 and 24 mg/dL, 19–22 mg/dL, 25–26 mg/dL in case of Menarini sensor and 6–11 mg/dL, 9–14 mg/dL, 12–18 mg/dL in case of Abbot sensor for 15, 30 and 60 min prediction horizon respectively. The results suggested that the proposed multimodal model achieved higher prediction accuracy compared to unimodal approaches; with upto 96.7% prediction accuracy; supporting its potential as a generalizable solution for interstitial glucose prediction and personalized management in the type 2 diabetes population.
Cavity-enhanced continuous-wave microscopy with potentially unstable cavity length
Abstract Microscopy gives access to spatially resolved dynamics in different systems, from biological cells to cold atoms. A big challenge is maximizing the information per used probe particle to limit the damage to the probed system. We present a cavity-enhanced continuous-wave microscopy approach that provides enhanced signal-to-noise ratios at fixed damage compared to standard single-pass microscopy. Employing a self-imaging 4f cavity, we show contrast enhancement for controlled test samples as well as biological samples. For thick samples, the imaging cavity leads to a new form of dark-field microscopy, where the separation of scattered and unscattered light is based on optical path length. We theoretically show that enhanced signal, signal-to-noise, and signal-to-noise per damage are also retrieved when the cavity is not length-stabilized. Our results provide an approach to cavity-enhanced microscopy with non-length-stabilized cavities and might be used to enhance the performance of dispersive imaging of ultracold atoms.
Can science address painters’ problems?
The role of selected NKG2DLs such as MICA, MICB and ULBP4 as potential markers in multiple sclerosis
Reserve optimization model of wind power with the coordination of multiple type electric power system sources
Author Correction: GelMA as scaffold material for epithelial cells to emulate the small intestinal microenvironment
Cell’s sugar coating mapped at below-nanometre resolution
A novel mineralization-inductive peptide derived from CEMP1functinal domains
Scientists mourn Tom Lehrer — nerdiest of singer-songwriters
Quantum-like nonlinear interferometry with frequency-engineered classical light
Daily briefing: The brain deploys immune cells at the mere sight of sickness
Family social support during incarceration: implications for health upon release
Abstract Incarceration is associated with adverse physical and mental health that are often brought to light during reentry into the community, particularly in the immediate period following release. Social support, specifically from family members, has been identified as a key protective factor that may promote health and reintegration success among formerly incarcerated individuals. However, less is known about how specific types of family support—emotional and instrumental—relate to health outcomes following release from incarceration. The current study uses data from 517 individuals incarcerated in a large Texas prison, surveyed before and approximately one month after release, to examine the relationship between family support and self-rated physical and mental health. Logistic regression models revealed that strong emotional family support was significantly associated with better self-rated physical health and mental health one month post-release. Additionally, strong instrumental family support predicted better mental health but not physical health among respondents. These findings highlight the crucial role of emotional and instrumental familial support systems in fostering and reducing health disparities and promoting equity among justice-impacted populations.
Sign language recognition based on dual-channel star-attention convolutional neural network
Effects of loading modes and temperature on fracture properties of limestone
Retraction for Goswami et al., A bifunctional tRNA import receptor from <i>Leishmania</i> mitochondria
Effect of virtual walk in green or urban spaces on pain perception among healthy adults
Predicting flavonoid physicochemical properties using topological indices and regression modeling
Phage-based delivery of CRISPR-associated transposases for targeted bacterial editing
Phage λ, a well-characterized temperate phage, has been recently leveraged for bacterial genome editing by selectively delivering base editors into targeted bacterial species. We extend this concept by engineering phage λ to deliver CRISPR-guided transposases, accomplishing large insertions and targeted gene disruptions. To achieve this, we engineered phage λ using homologous recombination paired with Cas13a-based counterselection for precise phage modifications. Initially, we established the utility of Cas13a in phage λ by conducting minimal recoding edits, deletions, and insertions. Subsequently, we scaled up the engineering to embed the comprehensive DNA-editing CRISPR-Cas transposase (DART) system within the phage genome, creating λ-DART phages. These modified λ-DART phages were then employed to infect Escherichia coli , generating CRISPR RNA-guided transposition events in the host genome. Applying our engineered λ-DART phages to monocultures and a mixed bacterial community comprising three genera led to efficient, precise, and specific gene knockouts and insertions in the targeted E. coli cells, achieving editing efficiencies surpassing 50% of the population. This research enhances phage-mediated genome editing by enabling efficient in situ gene integrations in bacteria, offering an avenue for further application in microbial community contexts. This scalable method enables flexible microbial genome editing in situ to manipulate the function and composition of diverse ecosystems.
The transcription factor WRKY25 can act as redox switch to drive the expression of WRKY53 during leaf senescence in Arabidopsis
Abstract Senescence requires high plasticity and, therefore, must be coordinated by a complex regulatory network. Notably, WRKY transcription factors highly impact senescence regulation. WRKYs can form homo- and heterodimers and contain the binding motifs of WRKY factors in their promoters already forming a complex regulatory network between themselves. For the Arabidopsis hub gene WRKY53 , WRKY18 acts as a strong negative while WRKY25 serves as strong positive regulator, creating a smaller subnetwork with high complexity, which we analyzed in detail. Activation of WRKY53 expression by WRKY25 is redox sensitive while repression by WRKY18 was not. Deletions and domain-swapping between WRKY18 and WRKY25 revealed that the N-terminal domain of WRKY25 is crucial for its activator effect on WRKY53 expression. Moreover, WRKY25 does not form homodimers but is able to heterodimerize with WRKY18 also requiring its N-terminal domain. The impact on senescence regulation and on WRKY53 expression was validated in planta using transgenic complementation lines of the wrky25 mutant. Modeling WRKY25 in silico indicated a putative covalent lysine-cysteine NOS redox switch. LC–MS analyses suggest that the NOS bridges really exist. We propose that WRKY25 acts as a redox sensor, balancing the expression and interactions of the WRKY53/WRKY25/WRKY18 network to ensure progressive senescence induction.