Bayesian Integration in Sense of Agency: Understanding Self-Attribution and Individual Differences

Abstract Genomic instability is a hallmark of cancer and contributes to disease progression through various mechanisms. DNA copy number variation (CNV) is one of the most significant consequences causing a phenomenon called aneuploidy. Aneuploidy is characterized by an imbalanced genome and is associated with lethal progression in prostate cancer. Currently, sequencing-based CNV analysis is widely used to identify cancer-specific aneuploidy that contribute to tumorigenesis, immune evasion, and disease progression. Performing low-pass whole genome sequencing (LP-WGS) on DNA extracted from bulk tissues or cell cultures, is a convenient approach to use for detecting aneuploidy and focal CNVs. However, conventional bulk LP-WGS methods average genomic signals across thousands of cells, often overlooking tumor heterogeneity and small populations with distinct CNV patterns. To address this limitation, we optimized a microfluid nanowell-based system to analyze aneuploidy patterns at a single-cell level using disassociated tissues and isolated nuclei from prostate samples. This pipeline is compatible with both preclinical models and clinical specimens and can be integrated with other multi-omics platforms to generate multimodal datasets for basic and translational research. By leveraging high-throughput single-cell sequencing technologies, our approach enables direct DNA CNV measurement with high sensitivity at a single-cell resolution. Ultimately, this approach aims to improve the resolution of CNV detection, enhance molecular stratification and enable multimodal data integration for better understanding of clinical prostate cancer samples. Citation Format: Chloe Springer, Faith Kim, Kathryn Echandía-Monroe, Aditi Shirke, Gustavo Guitierrez-Cruz, Madeline Wong, Kun-Lin Ho, Duanduan Ma, Elise G. DeArment, Amina Ali, William D. Figg, Gregory Chesnut, Matthew G. Vander Heiden, Xiaofeng A. Su. Characterizing aneuploid tumor heterogeneity in prostate cancer through single-cell DNA sequencing [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr A060.

This study presents DeepVisionAnalytics, an integrated framework that combines eye tracking, OpenCV-based computer vision (CV), and machine learning (ML) to support objective analysis of consumer behaviour in visually driven tasks. Unlike conventional self-reported surveys, which are prone to cognitive bias, recall errors, and social desirability effects, the proposed approach relies on direct behavioural measurements of visual attention. The system captures gaze distribution and fixation dynamics during interaction with products or interfaces. It uses AOI-level eye tracking metrics as the sole behavioural signal to infer candidate choice under constrained experimental conditions. In parallel, OpenCV and ML perform facial analysis to estimate demographic attributes (age, gender, and ethnicity). These attributes are collected independently and linked post hoc to gaze-derived outcomes. Demographics are not used as predictive features for choice inference. Instead, they are used as contextual metadata to support stratified, segment-level interpretation. Empirical results show that gaze-based inference closely reproduces observed choice distributions in short-horizon, visually driven tasks. Demographic estimates enable meaningful post hoc segmentation without affecting the decision mechanism. Together, these results show that multimodal integration can move beyond descriptive heatmaps. The platform produces reproducible decision-support artefacts, including AOI rankings, heatmaps, and segment-level summaries, grounded in objective behavioural data. By separating the decision signal (gaze) from contextual descriptors (demographics), this work contributes a reusable end-to-end platform for marketing and UX research. It supports choice inference under constrained conditions and segment-level interpretation without demographic priors in the decision mechanism.

Deep learning-based multimodal pathogenomics integration for precision cancer prognosis.

Nanozymes are enzyme-mimicking nanomaterials that are promising for diverse biomedical applications; they enable stable, tunable, and multifunctional cancer therapies via exploiting tumor microenvironment (TME) cues to regulate reactive oxygen species (ROS) and catalytic activities locally, aided by state-of-the-art fabrication methods and artificial intelligence (AI)-assisted designs. This review summarizes recent developments in nanozyme-based cancer therapy, focusing on the underlying catalytic mechanisms, material classifications, and their multimodality integration for cancer treatment. It further examines oxidase (OXD), peroxidase (POD), catalase (CAT), and superoxide dismutase (SOD)-like nanozymes in chemodynamic (CDT), photothermal (PTT), photodynamic (PDT), sonodynamic (SDT), immune and starvation therapies (ST), emphasizing single-atom, multimetallic, biomimetic, and AI-assisted design strategies of nanozymes. Single-atom and multimetallic nanozymes offer superior catalytic precision, atom efficiency, and programmable pathways over conventional nanomaterials. While AI-assisted design accelerates discovery of optimal compositions and therapeutic environment compatibility, enabling controlled ROS generation and TME responsiveness and hence enhancing tumor selectivity and therapeutic efficacy, their combination may represent a transformative direction for precision cancer therapy. Despite encouraging progress, challenges related to in vivo specificity, long-term biosafety, scalable synthesis, and clinical translation remain. Addressing these issues through interdisciplinary innovation will be critical for advancing next-generation intelligent nanozyme platforms toward clinical oncology.

We present our new Brain/MINDS 3D digital marmoset brain atlas version 2.0 (BMA2.0), a population-based 3D digital brain atlas of the common marmoset (Callithrix jacchus), designed to overcome the limitations of previous single subject atlases that are prone to structural biases arising from individual variation. Here, manually delineated cortical regions from 10 myelin-stained brains were used to create a generalized cortical parcellation. Newly refined subcortical regions from a previous atlas and a completely new cerebellum parcellation were also incorporated, resulting in a comprehensive whole brain parcellation for both hemispheres. To facilitate multimodal data analysis, the atlas package includes co-registered average templates for myelin and Nissl staining from the same individuals, ex vivo MRI T2 (91 individuals), and in vivo MRI T2 (446 individuals). Cortical flat maps and pial, cortical mid-thickness, and white matter surfaces are also provided. BMA2.0 provides a central brain space for multimodal data integration, spatial analysis, and comparative neuroscience. Standard formats and transformations are provided for easy integration into existing workflows and interoperability with existing atlases.

This systematic review outlines pivotal advancements in deep learning-driven protein structure prediction and design, focusing on four core models—AlphaFold, RoseTTAFold, RFDiffusion, and ProteinMPNN—developed by 2024 Nobel Laureates in Chemistry: David Baker, Demis Hassabis, and John Jumper. We analyze their technological iterations and collaborative design paradigms, emphasizing breakthroughs in atomic-level structural accuracy, functional protein engineering, and modeling multi-component biomolecular interactions. Key innovations include AlphaFold3's diffusion-based framework for unified biomolecular prediction, RoseTTAFold's three-track architecture integrating sequence and spatial constraints, RFDiffusion's denoising diffusion for de novo protein generation, and ProteinMPNN's inverse folding for sequence–structure co-optimization. Despite major progress in applications such as binder design, nanomaterials, and enzyme engineering, challenges persist in dynamic conformational sampling, multimodal data integration, and generalization to non-canonical targets. We propose future directions, including hybrid physics-AI frameworks and multimodal learning, to bridge gaps between computational design and functional validation in cellular environments.

Abstract Psychiatric disorders are complex, disabling conditions that continue to rely on subjective diagnostic criteria due to the absence of objective biological markers. Neuroradiology has become a critical discipline for examining the structural, functional, and biochemical underpinnings of these disorders through advanced brain imaging. This review synthesizes findings from five major psychiatric conditions including major depressive disorder, schizophrenia, autism spectrum disorder, obsessive-compulsive disorder, and generalized anxiety disorder and briefly discuss behavioral variant of Alzheimer’s disease, a variant with neuropsychological overlay, across multiple imaging modalities, including structural MRI, diffusion tensor imaging, functional MRI, magnetic resonance spectroscopy, functional near-infrared spectroscopy, and positron emission tomography. We present a comparative overview of cross-condition and modality-specific findings, highlighting converging disruptions in frontolimbic and temporoparietal circuits, alongside unique neurobiological features in each disorder. We also acknowledge key confounds such as medication effects, comorbidities, and methodological variability that limit direct transdiagnostic inference. We further discuss methodological limitations, emerging trends such as multimodal integration and machine learning, and future directions for translating imaging data into clinically meaningful biomarkers.

Decoding polyphenol-protein interactions with deep learning: From molecular mechanisms to food applications.

Introduction Olfactory dysfunction is common in the Alzheimer’s Disease continuum, and olfaction may be altered before clinical syndrome onset. The present study aimed at investigating the functional connectivity of the olfactory cortex and its correlation with olfaction performance in a group of patients with Mild Cognitive Impairment (MCI) who subsequently converted or not converted to Alzheimer’s Disease (AD) dementia. Methods At baseline, 30 MCI patients were evaluated with the Sniffin’ Sticks (threshold, discrimination, and identification) to assess olfactory capacities, and they were followed up over time to identify converter and stable patients. Resting-state fMRI data acquired at baseline were analyzed to assess functional connectivity of left and right olfactory cortex. Beta values were extracted from the stable versus converter contrasts and correlated with olfactory scores. Results Functional connectivity of the olfactory cortex was significantly increased with the posterior cingulate cortex, and significantly decreased with middle cingulate cortex, supplementary motor area, and left pre- and postcentral gyri, in converter compared to stable patients. Reduced negative functional connectivity between olfactory cortex and left angular gyrus emerged in converter patients, and a negative correlation was found between angular gyrus and discrimination scores. Discussion Our findings indicate alterations of functional connectivity of the olfactory cortex in subjects with MCI at risk of conversion to AD dementia, even at the early stages of the disease. Additionally, the negative correlation between olfactory ability and the angular gyrus functional connectivity, a cerebral region known to be involved in multisensory integration processing, may be considered as a marker of disease progression.

Understanding and classifying driving behavior is a critical component of modern intelligent transportation systems, with implications for traffic safety, fuel efficiency, and personalized driver support. As sensor-equipped mobile devices become increasingly pervasive, new opportunities have emerged for implementing data-driven behavior recognition systems in a cost-effective and accessible manner. This study presents a comprehensive and low-cost mobile framework for classifying driving behaviors using data collected entirely via a smartphone application. Unlike prior approaches that rely on embedded hardware, the proposed system performs all data acquisition and recording through a standard smartphone paired with a bluetooth-based on-board diagnostics II (OBD-II) adapter. The framework integrates multimodal sensor sources, including engine control unit (ECU) data, inertial motion sensors, enriched road metadata via the Overpass application programming interface (API), and environmental audio signals. Rather than isolating a single data domain, the system unifies mechanical, contextual, and behavioral dimensions to enable robust driving style analysis. Driving behavior was categorized into three classes (calm, normal, aggressive) using sliding time windows of 3, 5, 7, and 9 s. The effects of both window duration and data source composition on model performance were thoroughly evaluated. Classical machine learning models (artificial neural network (ANN), support vector machine (SVM), logistic regression (LR), Naive Bayes (NB)) based on engineered features were compared against deep learning architectures (convolutional neural network (CNN), long short-term memory (LSTM), gated recurrent unit (GRU), recurrent neural network (RNN)) trained on raw multivariate sequences. Results showed that multimodal integration substantially improved classification accuracy, with CNN achieving the highest performance. Additionally, the study incorporated patch-based time series transformer (PatchTST), a modern transformer-based architecture designed for time-series classification, across 128 experimental configurations. While CNN remained the top performer in overall accuracy, PatchTST yielded consistently stable and competitive results, particularly in long-window and feature-rich settings. Importantly, statistical analyses confirmed the significance of differences across feature sets, time windows, and model types. This included architectural parameters such as model depth and latent dimensionality, evaluated through analysis of variance (ANOVA) and post-hoc Tukey’s honestly significant difference (HSD) tests. By enabling high-accuracy driving behavior classification using only smartphone-based sensing, this study contributes a practical and scalable solution. The inclusion of attention-based PatchTST modeling further extends the methodological breadth, highlighting the role of transformer architectures in multivariate time-series analysis. Collectively, these contributions underscore the feasibility of deploying robust and intelligent driver monitoring systems in real-world environments.

The expanding diversity of synthetic chemicals is increasing ecological risk, yet many predictive models rely on single-species data and inadequately capture interspecies variability in sensitivity. We propose a generalized toxicity prediction framework (GTGT) that predicts species-resolved acute toxicity (log10LC50; LC50 in mg/L) from chemical descriptors, exposure duration, and species features─taxonomy embeddings and mitochondrial Cytochrome b (cytb) sequence embeddings─within a unified deep-learning architecture. Using a dataset of 2860 compounds and 297 fish species, GTGT outperformed representative state-of-the-art models, achieving external-test R2 = 0.83 and RMSE = 0.49. Ablation analyses show that chemical and exposure descriptors provide baseline performance, whereas biological features are critical to capture interspecies susceptibility. Comparative analyses further indicate that taxonomy embeddings encode hierarchical evolutionary relationships, while cytb sequences capture molecular divergence, providing complementary information for robust cross-species prediction. We also provide a web platform for single- and multicompound predictions across multiple fish species, enabling model-based species sensitivity distribution (SSD) curves. This framework links chemical, biological, and exposure dimensions to support SSD parametrization and derivation of protective thresholds for comparison with environmentally relevant exposures, rather than serving as a direct risk metric.

Aging commonly leads to balance problems, yet the neural processes driving this decline remain unclear. Recent structural, resting-state, and EEG combined with virtual reality (VR) studies suggest that age-related instability stems from reduced flexibility in combining visual, vestibular, and somatosensory cues rather than from losses in any single system. These findings indicate that diminished neural adaptability is a key contributor to balance impairment and point toward specific network-level mechanisms that future interventions may target.

Comparative and exploratory study of ATR and diffuse reflectance mid-infrared spectroscopy for coal property analysis.

Artificial Intelligence in Snoring Sound Analysis: OSA Detection and Obstruction Site Classification, a Systematic Review.

Advancing precision psychiatry: Machine learning integration with neuroimaging for early detection and diagnosis of Obsessive-Compulsive Disorder.

BackgroundThe monocyte substate MS1, identified through single-cell RNA sequencing (scRNA-seq), is enriched in sepsis and shares a transcriptional profile with monocytic myeloid-derived suppressor cells (M-MDSCs), which inhibit T cell activation and cytotoxicity in cancer and chronic inflammation. M-MDSCs can be induced by a range of signals, including inflammatory cytokines such as IL-6 and IL-1β, as well as growth factors like GM-CSF and M-CSF involved in emergency myelopoiesis. While MS1 expansion has been observed in sepsis, the cytokine drivers of this response remain incompletely defined.Figure 1.IL-6 levels parallel the early expansion and subsequent decline of an immunosuppressive monocyte state in sepsis.(a) MS1 monocyte fractional abundance within PBMCs decreases progressively from presentation through recovery.(b) IL-6 plasma concentrations follow a similar temporal trajectory.Samples per timepoint: Day 0 – sepsis (n=67), sterile inflammation (n=26); Day 1 – sepsis (n=60), sterile inflammation (n=23); Day 3 – sepsis (n=53), sterile inflammation (n=15); Day 7 – sepsis (n=27), sterile inflammation (n=3); Convalescence (Day 28+) – sepsis (n=14), sterile inflammation (n=3). Points and lines represent individual subjects, colored by phenotype; boxes show median and interquartile range. IL-6 concentrations (pg/mL) were measured in matched plasma samples using multiplex proteomics. Significance was assessed using Wilcoxon rank-sum test with Benjamini-Hochberg correction. Statistical comparisons between sepsis and sterile inflammation (SI) are shown within the plotting area; comparisons versus healthy controls (HC) are shown outside the plotting area. *p < 0.05, **p < 0.01, ***p < 0.001.MethodsWe performed scRNA-seq with paired surface protein profiling on peripheral blood mononuclear cells (PBMCs) from patients with sepsis and sterile inflammation, sampled at presentation to the emergency department and longitudinally through recovery. In total, we analyzed samples from 130 individuals: 71 with sepsis, 27 with sterile inflammation, and 12 healthy controls recruited outside the ED. After quality control, 560,867 high-quality single cells were analyzed using multimodal integration. Matched plasma samples were profiled using a high-plex cytokine proteomics platform to assess candidate inducers of the MS1 state.ResultsMS1 monocytes were elevated at sepsis onset and declined progressively through day 7, remaining low into convalescence (Fig. 1a). Among candidate cytokines, only IL-6 exhibited a parallel temporal decline (Fig. 1b). Spearman correlations between IL-6 levels and MS1 abundance ranged from ρ=0.39 at day 0 to ρ=0.69 at convalescence. Other cytokines, including IL-1β, IL-8, TNF-α, and IL-10, showed less consistent trajectories and weaker correlations.ConclusionThese findings demonstrate an association between IL-6 and early expansion of an immunosuppressive monocyte state during sepsis, suggesting a putative IL-6–dependent emergency myelopoiesis program. The observed temporal relationship may reflect an early, IL-6–driven adaptation that transitions toward immune restoration during recovery. Defining cytokine–cell state relationships in acute infection may inform strategies to modulate myeloid cell dysregulation in patients with sepsis.DisclosuresAll Authors: No reported disclosures

Optical coherence tomography (OCT), with its advantages of non-invasiveness, non-contact, rapid imaging, and high resolution, has become an indispensable core imaging tool in the diagnosis and treatment of fundus diseases. It provides clinicians with critical structural insights for accurately assessing lesion characteristics and severity, dynamically monitoring disease progression and treatment efficacy, and optimizing therapeutic strategies. However, in the era of precision medicine, the application effectiveness of OCT still faces significant challenges. Deeply understanding and acknowledging its limitations is fundamental to maximizing the value of OCT and seeking effective solutions. This article focuses on the key bottlenecks and breakthrough paths of OCT in the diagnosis and treatment of fundus diseases, systematically analyzing the core issues from three dimensions, including the physical constraints of the device itself, the physiological and pathological variations among individual patients, and the subjectivity of physician interpretation. Closely following the cutting-edge trends and development opportunities in the OCT field, the counter strategies, like continuous technological innovation, deep integration of artificial intelligence, multimodal imaging integration, optimization of patient workflows and protocols, and standardized physician training, are also discussed. The aim is to provide clinicians with a comprehensive insight into the current status and future directions of OCT applications, helping them better understand existing challenges, more effectively utilize current technologies, and grasp the opportunities brought by technological innovation and diagnostic model transformation, and thus promoting the continuous improvement of precision diagnosis and treatment levels for fundus diseases.

Foundation model embeddings for multimodal oncology data integration.

Pipeline infrastructure constitutes the primary transportation system within the oil and gas industry, where operational safety is critically dependent on advanced in-line inspection technologies. This study presents a comprehensive analysis of eddy current testing (ECT) applications for pipeline integrity assessment. The fundamental principles of ECT are first elucidated, followed by a systematic comparative evaluation of five key ECT methodologies: conventional, multi-frequency, remote field, pulsed, and array eddy current techniques. The analysis examines their detection mechanisms, technical specifications, comparative advantages, and current developmental trajectories, with particular emphasis on future technological evolution. Subsequently, integrating global pipeline infrastructure development trends and market requirements, representative designs of pipeline inspection tools are detailed and we review relevant industry applications. Finally, persistent challenges in ECT applications are identified, particularly regarding adaptability to complex operational environments, quantification accuracy for micro-scale defects, and predictive capability for defect progression. This study proposes that future ECT equipment development should prioritize multi-modal integration, miniaturization, and intelligent analysis to enable comprehensive pipeline safety management throughout the entire asset lifecycle.