Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees [new] ...achieved for multistable dynamics (Morse-Smale, cont attractors) ensuring $\varepsilon$-$δ$ closeness for infinite time, and a temporal generalization bound.

A Dynamical Microscope for Multivariate Oscillatory Signals: Validating Regime Recovery on Shared Manifolds [new] recovers dynamical laws by converting signals to latent trajectory geometry and flow field metrics, discerning overlapping regimes.

Importance inversion transfer identifies shared principles for cross-domain learning [new] ...by prioritizing invariant structural anchors over idiosyncratic features to generalize across diverse biological, linguistic, molecular, and social networks.

STRAND: Sequence-Conditioned Transport for Single-Cell Perturbations [new] ...predicts single-cell transcriptional responses by conditioning a transport process on regulatory DNA sequence at the perturbation locus for zero-shot inference.

Beyond SMILES: Evaluating Agentic Systems for Drug Discovery [new] reveals five capability gaps in generalizing across diverse drug discovery tasks, particularly peptides and in vivo data, due to architectural limitations.

Fully-automated sleep staging: multicenter validation of a generalizable deep neural network for Parkinson's disease and isolated REM sleep behavior disorder [new]

GeneReL: A Large Language Model-Powered Platform for Gene Regulatory Relationship Extraction with Community Curation [new] ... uses tiered LLM pipeline & novel gene norm to extract plant reg data, fostering comm valid. for unique DB.

Validating Interpretability in siRNA Efficacy Prediction: A Perturbation-Based, Dataset-Aware Protocol [new] Perturbation proto tests high-sal. edits' impact predictions vs. controls, uncovering failure modes & cross-dataset collapse. Validation essential.

EVA: Towards a universal model of the immune system [new] A multimodal, cross-species foundation model harmonizing transcriptomics & histology for unified patient representations in immune-mediated diseases.

ICODEN: Ordinary Differential Equation Neural Networks for Interval-Censored Data [new] ...flexibly models hazard functions via deep neural networks, deriving cumulative hazard by solving ODEs for assumption-lean survival analysis.

Towards Universal Spatial Transcriptomics Super-Resolution: A Generalist Physically Consistent Flow Matching Framework [new] Decouples gene semantics from spatial geometry via SSL and uses flow matching for optimal transport, ensuring local mass conservation.

Large-scale quantum computing framework enhances drug discovery in multiple stages [new] ...by developing a stable 2000-node Coherent Ising Machine for a CADD workflow, identifying and validating novel druggable sites and compounds.

A machine learning approach to identify key Epigenetic Transcripts for Ageing research in human blood (Epitage) [new] identifies 48 age-corr transcripts (13 genes) in blood,highlighting novel markers. An R package aids validation.

DIA-CLIP: a universal representation learning framework for zero-shot DIA proteomics [new] achieves zero-shot PSM inference by aligning peptides & spectra into a unified latent space, eliminates run-specific training requirements.

Metadiffusion: inference-time meta-energy biasing of biomolecular diffusion models [new] Steers PT diffusion models via grad-guided denoising to generate diverse biomol. conformations beyond compact DB biases, without retraining.

A spectral framework for measuring diversity in multiple sequence alignments [new] Framework defines diversity as the no. of independent alignment positions reproducing observed seq variation, quantifying evolutionary constraints.

Automated Detection of Quiet and Non-Quiet Sleep in Preterm Neonates from aEEG: Towards Predicting Brain Maturation [new] Auto sleep det. via simple monitor predicts brain matur., enabling obj. early impair. ID in preterm neonates.

Multiplexed high-content imaging uncovers morphological diversity of lymphocyte activation and dysfunction [new] Assessing SC morph, organelle arch, & activation states across perturbations reveals morph diversity in lymphocyte A/D.

Profiling co-occurrent morphological phenotypes and their degree of expression severity in vacuolated cells by holo-tomographic flow cytometry and fractal analysis [new]

BRIDGE: Biological Antimicrobial Resistance Inference viaDomain-Knowledge Graph Embeddings [new] infers novel AMR gene links by integrating gene nbhd and protein interact networks into a comprehensive knowledge graph for prediction.

A multi-component power-law penalty corrects distance bias in single-cell co-accessibility and deep-learning chromatin interaction predictions [updated]

Using AI responsibly in scientific publishing [new] Discusses journal policies and responsible application of the technology in writing, peer review, and publishing.

Diffusion model generating regulatory DNAs [new] ...to program desired genetic functions.

VC-RDAgent: An efficient rare disease diagnosis agent via virtual case construction informed by hybrid statistical-metric and hyperbolic-semantic prioritization [new]

What matters in building vision–language–action models for generalist robots](htw] Comparative analysis reveals key factors for models' utility in generalist robotics.

When large language models are reliable for judging empathic communication [new] ...their empathic communication assessments hinge on clear, specific evaluation criteria.

ATLAS: Population-Level Disease Locus Discovery via Differential Attention in Genomic Language Models [new] identifies disease genes & loci directly from raw sequences via gLM attns signals, bypassing variant calls & sup training.

A Systematic Benchmark for Peptide Property Prediction [new] standardizes 15 datasets & unifies evaluation for realistic model generalization, overcoming data heterogeneity & inconsistent standards.

Latent-centric Isotropic Resolution Enhancement for Expansion Microscopy Imaging via Neural Compression and Self-supervised Learning [new]

Importance inversion transfer identifies shared principles for cross-domain learning [new] ...prioritizing domain-invariant structural anchors over idiosyncratic features, formalizing a framework for robust knowl. transfer across heterogeneous, noisy systems.

Drug Release Modeling using Physics-Informed Neural Networks [new] ...integrates Fick's law with limited experimental data for accurate long-term predictions from short-term drug release measurements in various film types.

Fully-automated sleep staging: multicenter validation of a generalizable deep neural network for Parkinson's disease and isolated REM sleep behavior disorder [new]

IntelliFold-2: Surpassing AlphaFold 3 via Architectural Refinement and Structural Consistency [new] Arch. refinements & multiscale structural consistency enhance OS biomol. struct. pred. via novel attention, sampling, & loss methods.

scBench: Evaluating AI Agents on Single-Cell RNA-seq Analysis [new] via a benchmark of 394 verifiable problems from practical scRNA-seq workflows, assessing AI agents' ability to extract biological insight.

AntigenLM: Structure-Aware DNA Language Modeling for Influenza [new] ...uses pretraining on intact influenza genomic functional units to capture evolutionary constraints, enabling generalization for antigen evolution prediction.

Reward-Guided Discrete Diffusion via Clean-Sample Markov Chain for Molecule and Biological Sequence Design [new] bypasses noisy intermed. rewards by forming a Metropolis-Hastings Markov chain of clean samples using sequential forward/backward diffusion props.

Optimizing Protein Tokenization: Reduced Amino Acid Alphabets for Efficient and Accurate Protein Language Models [new] ... combining reduced AA alphabets & subword tokenization for shorter inputs & effective pLM tokenization.

CoPrimeEEG: CRT-Guided Dual-Branch Reconstruction from Co-Prime Sub-Nyquist EEG [new] Neural framework reconstructs hi-rate EEG from co-prime sub-Nyquist low-rate data via CRT-guided dual-branch encoder, predicting masks & bandpower.

DynMoCo: a Novel AI Framework to Reveal Modular Substructures of Protein From Molecular Dynamics [new] identifies time-evolving, coherent protein substructures via dynamic community detection on mol. graphs from MD simulations.

Autoregressive forecasting of future single-cell state transitions [new] ...by training a generative AI to autoregressively predict ordered semantic cell codes from static RNA-seq snaps, yielding long-range trajectories & landscapes.

Leveraging Foundation Models for the Characterisation of Small RNA Properties [new] ...compares endogenous (miRNAs, piRNAs) and synthetic siRNAs, revealing class-specific features to understand RNA function and improve siRNA design.

Parsimonious cell co-localization scoring for spatial transcriptomics [new] quantifies cell type co-occurrence using a per-cell metric based on neighborhood proportion products to map hotspots & multivariate niches.

LineageSim: A Single-Cell Lineage Simulator with Fate-Aware Gene Expression [new] generates single-cell lineage data where prog. gene expr. encodes latent signals of future cell fates enabling benchmarks for fate predict. algorithms.

PRIZM: Combining Low-N Data and Zero-shot Models to Design Enhanced Protein Variants [new] selects optimal zero-shot models using minimal experimental data, then ranks in silico variant libraries for targeted testing.

A multi-agent platform for assessment and improvement of bioinformatics software documentation [new] iteratively refines documentation as a testable object. Agents emulate user interact, assess against stand task criteria to correct.

Integration and harmonization of cell shape images for generative modeling [new] ...to unify diverse 2D/3D cell image data, creating synthetic cells for augmented ML analysis of cellular responses to mechanical cues.

Improving Protein Structure Prediction Using Integrative Cryo-EM and Ion Mobility Mass Spectrometry Modeling [new] ...by combining low-res cryo-EM dens and IM-MS CCS as restraints in a Rosetta scoring function for monomeric proteins.

Reading TEA leaves for de novo protein design [new] ...by using a protein language model embedding-derived alphabet for fast Metropolis sampling, enabling template-guided and unconditional protein sequence design sans homologs.

Token Alignment for Verifying LLM-Extracted Text [new] efficiently verifies LLM extractions by aligning discontiguous text via subword tokenization and ordered alignment, using new datasets.

A federated graph learning method to realize multi-party collaboration for molecular discovery [new] uses Lanczos projection to aggregate diverse molecular data.