ArXiv最新文献

Millions of children worldwide are affected by severe rare Mendelian disorders, yet exome and genome sequencing still fail to provide a definitive molecular diagnosis for a large fraction of patients, prolonging the diagnostic odyssey. Bridging this gap increasingly requires transitioning from DNA-only interpretation to multi-modal diagnostic reasoning that combines genomic data, transcriptomic sequencing (RNA-seq), and phenotype information; however, computational frameworks that coherently integrate these signals remain limited. Here we present RareCollab, an agentic diagnostic framework that pairs a stable quantitative Diagnostic Engine with Large Language Model (LLM)-based specialist modules that produce high-resolution, interpretable assessments from transcriptomic signals, phenotypes, variant databases, and the literature to prioritize potential diagnostic variants. In a rigorously curated benchmark of Undiagnosed Diseases Network (UDN) patients with paired genomic and transcriptomic data, RareCollab achieved 77% top-5 diagnostic accuracy and improved top-1 to top-5 accuracy by ~20% over widely used variant-prioritization approaches. RareCollab illustrates how modular artificial intelligence (AI) can operationalize multi-modal evidence for accurate, scalable rare disease diagnosis, offering a promising path toward reducing the diagnostic odyssey for affected families.

Many modern ultrasound beamformers report improved image quality when evaluated using classical criteria like the contrast ratio and contrast-to-noise ratio, which are based on summary statistics of regions of interest (ROIs). However, nonlinear beamformers and post-processing methods can substantially alter these statistics, raising concerns that the reported improvements may reflect changes in dynamic range or remapping rather than a reflection of true information gain, such as clutter suppression. New criteria like the generalized contrast-to-noise ratio (gCNR) address these concerns, but rely on noisy estimates of the underlying distribution. To address this, we introduce a new image quality criterion, called the contrast order (CO), defined as the expected value of the sign of the difference in brightness between two ROIs. The CO is invariant under all strictly monotonic transformations of the image values, as it depends only on their relative ordering, and is interpretable as the probability that one ROI is brighter than the other minus the probability that it is darker. Unlike the gCNR, the CO has a simple unbiased estimator whose variance decreases with the number of samples in each ROI. We further propose the effective contrast ratio (ECR), which calibrates the contrast order to the familiar contrast ratio such that the two coincide under ideal Rayleigh-speckle statistics. Together, the CO and ECR provide order- and sign-preserving, dynamic-range-invariant criteria for evaluating lesion contrast, offering a principled alternative to classical and newer image quality criteria when assessing modern beamformers.

We introduce a reproducible, bias-resistant machine learning framework that integrates domain-informed feature engineering, nested cross-validation, and calibrated decision-threshold optimization for small-sample neuroimaging data. Conventional cross-validation frameworks that reuse the same folds for both model selection and performance estimation yield optimistically biased results, limiting reproducibility and generalization. Demonstrated on a high-dimensional structural MRI dataset of deep brain stimulation cognitive outcomes, the framework achieved a nested-CV balanced accuracy of 0.660,$pm$,0.068 using a compact, interpretable subset selected via importance-guided ranking. By combining interpretability and unbiased evaluation, this work provides a generalizable computational blueprint for reliable machine learning in data-limited biomedical domains.

Spatiotemporal flows of neural activity, such as traveling waves, have been observed throughout the brain since the earliest recordings; yet there is still little consensus on their functional role. Recent experiments and models have linked traveling waves to visual and physical motion, but these observations have been difficult to reconcile with standard accounts of topographically organized selectivity and feedforward receptive fields. Here, we introduce a theoretical framework that formalizes and generalizes the connection between 'motion' and flowing neural dynamics in the language of equivariant neural network theory. We consider 'motion' not only in physical or visual spaces, but also in more abstract representational spaces, and we argue that recurrent traveling-wave-like dynamics are not just useful but necessary for accurate and stable processing of any signal undergoing such motion. Formally, we show that for any non-trivial recurrent neural network to process a sequence undergoing a flow transformation (such as visual motion) in a structured equivariant manner, its hidden state dynamics must actively realize a homomorphic representation of the same flow through recurrent connectivity. In this ''spatiotemporal perspective on dynamical computation'', traveling waves and related flows are best understood as faithful dynamic representations of stimulus flows; and consequently the natural inclination of biological systems towards such dynamics may be viewed as an innate inductive bias towards efficiency and generalization in the spatiotemporally-structured dynamical world they inhabit.

Detecting spatial patterns is fundamental to scientific discovery, yet current methods lack statistical consensus and face computational barriers when applied to large-scale spatial omics datasets. We unify major approaches through a single quadratic form and derive general consistency conditions. We reveal that several widely used methods, including Moran's I, are inconsistent, and propose scalable corrections. The resulting test enables robust pattern detection across millions of spatial locations and single-cell lineage-tracing datasets.

We introduce Multi-scale Adaptive Recurrent Biomedical Linear-time Encoder (MARBLE), the first textit{purely Mamba-based} multi-state multiple instance learning (MIL) framework for whole-slide image (WSI) analysis. MARBLE processes multiple magnification levels in parallel and integrates coarse-to-fine reasoning within a linear-time state-space model, efficiently capturing cross-scale dependencies with minimal parameter overhead. WSI analysis remains challenging due to gigapixel resolutions and hierarchical magnifications, while existing MIL methods typically operate at a single scale and transformer-based approaches suffer from quadratic attention costs. By coupling parallel multi-scale processing with linear-time sequence modeling, MARBLE provides a scalable and modular alternative to attention-based architectures. Experiments on five public datasets show improvements of up to textbf{6.9%} in AUC, textbf{20.3%} in accuracy, and textbf{2.3%} in C-index, establishing MARBLE as an efficient and generalizable framework for multi-scale WSI analysis.

High-resolution spatial transcriptomics platforms, such as Xenium, generate single-cell images that capture both molecular and spatial context, but their extremely high dimensionality poses major challenges for representation learning and clustering. In this study, we analyze data from the Xenium platform, which captures high-resolution images of tumor microarray (TMA) tissues and converts them into cell-by-gene matrices suitable for computational analysis. We benchmark and extend nonnegative matrix factorization (NMF) for spatial transcriptomics by introducing two spatially regularized variants. First, we propose Spatial NMF (SNMF), a lightweight baseline that enforces local spatial smoothness by diffusing each cell's NMF factor vector over its spatial neighborhood. Second, we introduce Hybrid Spatial NMF (hSNMF), which performs spatially regularized NMF followed by Leiden clustering on a hybrid adjacency that integrates spatial proximity (via a contact-radius graph) and transcriptomic similarity through a tunable mixing parameter alpha. Evaluated on a cholangiocarcinoma dataset, SNMF and hSNMF achieve markedly improved spatial compactness (CHAOS < 0.004, Moran's I > 0.96), greater cluster separability (Silhouette > 0.12, DBI < 1.8), and higher biological coherence (CMC and enrichment) compared to other spatial baselines. Availability and implementation: https://github.com/ishtyaqmahmud/hSNMF.

The BioModels database is one of the premier databases for computational models in systems biology. The database contains over 1000 curated models and an even larger number of non-curated models. All the models are stored in the machine-readable format, SBML. Although SBML can be translated into the human readable Antimony format, analyzing the models can still be time consuming. In order to bridge this gap, a LLM (large language model) assistant was created to analyze the BioModels and allow interaction between the user and the model using natural language. By doing so, a user can easily and rapidly extract the salient points in a given model. Our analysis workflow involved 'chunking' BioModels and converting them to plain text using llama3, and then embedding them in a ChromaDB database. The user-provided query was also embedded, and a similarity search was performed between the query and the BioModels in ChromaDB to extract the most relevant BioModels. The BioModels were then used as context to create the most accurate output in the chat between the user and the LLM. This approach greatly minimized the chance of hallucination and kept the LLM focused on the problem at hand. We illustrate the utility of this approach with a number of examples. The code is available at https://github.com/TheBobBob/BioModelsRAG. The website implementation is available at https://biomodelsrag.streamlit.app/.

In the last years in-vivo tractography has assumed an important role in neurosciences, for both research and clinical applications such as non-invasive investigation of brain connectivity and presurgical planning in neurosurgery. In more recent years there has been a growing interest in the applications of diffusion tractography for target identification in functional neurological disorders for an increasingly tailored approach. The growing diffusion of well-established neurosurgical procedures, such as deep brain stimulation or trans-cranial Magnetic Resonance-guided Focused Ultrasound, favored this trend. Tractography can indeed provide more accurate, patient-specific, information about the targeted region if compared to stereotactic atlases. On the other hand, this tractography-based approach is not very physician-friendly, and its heavily time consuming since it needs several hours for Magnetic Resonance Imaging data processing. In this study we propose a novel open-source deep learning framework called DeLTA-BIT (acronym of Deep-learning Local TrActography for BraIn Targeting) for fast target predictions, based on probabilistic tractography. The proposed framework exploits a convolutional neural network (CNN) to predict the location of the Ventral Intermediate Nucleus of the thalamus (VIM). The CNN was trained on the Human Connectome Project (HCP) dataset. The model capability in predicting the VIM location was tested both on the HCP (internal validation) and clinical data (external validation). Results from the internal validation have shown good capability in predicting the VIM region (mean DSC = 0.62+- 0.15, mean sDSC=0.76+- 0.17) by using just T1 images as input, in a time scale of fraction of second per subject. As for the clinical data, results have been compared with an atlas-based method demonstrating similar performance, but within a significantly shorter timeframe.

Foundation models have shown remarkable success in fitting biological visual systems; however, their black-box nature inherently limits their utility for understanding brain function. Here, we peek inside a SOTA foundation model of neural activity (Wang et al., 2025) as a physiologist might, characterizing each 'neuron' based on its temporal response properties to parametric stimuli. We analyze how different stimuli are represented in neural activity space by building decoding manifolds, and we analyze how different neurons are represented in stimulus-response space by building neural encoding manifolds. We find that the different processing stages of the model (i.e., the feedforward encoder, recurrent, and readout modules) each exhibit qualitatively different representational structures in these manifolds. The recurrent module shows a jump in capabilities over the encoder module by "pushing apart" the representations of different temporal stimulus patterns. Our "tubularity" metric quantifies this stimulus-dependent development of neural activity as biologically plausible. The readout module achieves high fidelity by using numerous specialized feature maps rather than biologically plausible mechanisms. Overall, this study provides a window into the inner workings of a prominent neural foundation model, gaining insights into the biological relevance of its internals through the novel analysis of its neurons' joint temporal response patterns. Our findings suggest design changes that could bring neural foundation models into closer alignment with biological systems: introducing recurrence in early encoder stages, and constraining features in the readout module.