1498 papers
This paper introduces BEAM-1, a biophysics-informed multiscale mathematical framework that integrates Fick's and Einstein-Stokes laws to predict and rank the blood-brain barrier permeability of 25 CNS drugs, successfully identifying Levodopa and Temozolomide as the most permeable while offering a cost-effective alternative to traditional in vitro testing.
This study employs an integrated bioinformatics approach to characterize the hypothetical protein RvY_05612 from *Ramazzottius varieornatus*, identifying it as a 98-amino-acid, basic, transmembrane Sec61β family protein localized to the endoplasmic reticulum that likely facilitates intracellular protein transport.
This study reveals that Citrus bark cracking viroid-induced stunting in hop is driven by a persistent, coordinated suppression of photosynthesis and nitrogen acquisition that begins in spring and leads to irreversible biomass loss, rather than a dominant immune-signaling response.
This study introduces a fully automated "virtual doctor" that diagnoses ADHD from standard 128-channel EEG recordings without expert intervention by utilizing a deep learning pipeline to identify diagnostically significant neurophysiological patterns—specifically in the 150-300 ms window—that traditional statistical methods overlook, thereby eliminating the human-expert bottleneck in EEG interpretation.
This study establishes a reproducible machine learning-assisted SHG microscopy workflow that reveals distinct collagen microarchitectures between fibrosis and trabecular bone in bone marrow biopsies, demonstrating that specific annotation strategies are critical for accurate quantitative morphometric analysis of tissue remodeling.
This study demonstrates that the SARS-CoV-2 main protease (Mpro) has remained evolutionarily constrained by structural and functional requirements since the Omicron era, preserving its substrate recognition and supporting its viability as a robust, long-term antiviral target.
This paper systematically investigates the manifold hypothesis in protein language models (pLMs) using a comprehensive battery of geometric and topological diagnostics, concluding that unlike single-cell foundation models, pLM representations are predominantly high-dimensional and linear rather than organized around low-dimensional manifolds.
This empirical study reveals that while biological large language models (SCFMs) do encode a real, low-dimensional, and partially linearized geometric structure for biological knowledge, this structure is modest, often nonlinearly entangled, and largely overlaps with what classical expression-based methods like PCA can already recover, falling short of the crisp, regular geometry observed in natural language models.
This paper presents a unified benchmark of eleven virtual cell models, revealing that while deep learning approaches significantly outperform baselines in representation tasks and combinatorial perturbation prediction, they generally fail to surpass simple linear models in predicting unseen single-gene perturbations, suggesting the field has achieved a "virtual microscope" for cell state analysis rather than a true "virtual simulator" for causal perturbation dynamics.
This study demonstrates that while sparse autoencoders effectively extract monosemantic, biologically interpretable features like sequence motifs and reading frames from DNA foundation models, they reveal a significant scarcity of features encoding complex regulatory logic, suggesting current models capture a genomic dictionary rather than a grammar engine.