deep learning

The Reasoning Paradox in Scientific Sandboxes: Why Extended Reasoning Fails in Autonomous Bioinformatics and How to Fix It.

This paper identifies two failure modes in extended reasoning models within bioinformatics tasks and introduces DTDR to improve accuracy and reduce errors.

PHATE-Derived Ancestry Coordinates Capture Continuous Population Structure in Human Genomic Data

This paper shows PHATE can better capture continuous population structure in genomic data than PCA or UMAP by tuning scale and balancing populations.

TileScope quantifies spatial gene expression relationships beyond colocalization

TileScope quantifies spatial gene expression relationships beyond simple colocalization, distinguishing regional enrichment from internal covariation in glioblastoma and other tissues.

Res ipSAE adhuc loquuntur: What the directionality of predicted aligned error in cofolding models can tell us about scoring diverse molecular complexes

This paper introduces ipSAE, a directional scoring framework for assessing confidence in predicted biomolecular complexes, showing improved performance over existing methods in protein-ligand and protein-DNA interactions.

BundleMorphometry: Spatially Resolved White Matter Bundle Shape Analysis Using Nonlinear Streamline-based Registration

BundleMorphometry quantifies local white matter bundle shape using nonlinear registration, revealing deformation patterns in glioma and Parkinson's disease datasets.

Parameter-Efficient Adaptation of Protein Language Models for Dual-Target Antimicrobial Peptide Discovery Under Strict Homology Constraints

This paper develops a parameter-efficient transfer learning framework to discover antimicrobial peptides with high potency and low cytotoxicity using protein language models.

scSTEC: Annotating Chromatin Subcompartments from Individual Single-Cell Hi-C Data via Direct Two-Dimensional Embedding and Chromosome-wise Consensus

scSTEC directly annotates chromatin subcompartments from individual single-cell Hi-C matrices without imputation or cross-cell sharing, achieving results comparable to scGHOST with greater interpretability.

Fast rate coding in heterogeneous spiking networks

This paper shows that heterogeneous spiking networks can produce fast rate-based dynamics without assuming Poisson spikes or slow dynamics.

Path knockout: an analysis tool for connectome-driven circuit discovery

Path knockout identifies dominant multi-synaptic pathways in connectomes by quantifying the impact of neuron removal on connection strength, outperforming existing methods in Drosophila circuits.

PhenoMe: An interactive AI framework for interpretable, microscopy-driven phenotyping

PhenoMe provides an open-source framework for analyzing microscopy images using foundation model-derived embeddings and experimental metadata to enable phenotyping and data exploration.

Particle picking and centering in cryo-electron micrographs via a vanishing radial kernel

The paper introduces a training-free method for particle picking and centering in cryo-EM by using a radial kernel convolution, achieving comparable reconstruction resolutions to reference-based methods.

Resolving how genetic perturbations act through gene regulatory paths

REGA-perturb resolves gene regulatory paths from genetic perturbations using reference epigenomes and RNA measurements, outperforming existing methods in identifying TF-target relationships and predicting regulatory changes.

GSET: Gene Signals from Embedding Trajectories select compact contexts for mechanistic GRN inference

GSET retrieves dynamic gene signals to improve mechanistic GRN inference by identifying bridge genes involved in cell-state transitions.

NeuralFieldManifold: Uncovering the Manifold Geometry of Local Field Potentials

This paper introduces a geometric framework for analyzing local field potentials (LFPs) by modeling their dynamics as low-dimensional K-tori and validates it across neural datasets.

ARCADE disentangles cell abundance from continuous transcriptional state in spatial transcriptomics

ARCADE improves spatial transcriptomic analysis by separating cell abundance from transcriptional state, enhancing deconvolution accuracy and interpretability in complex tissues.

Scalable Contact-Gated Stochastic Operators and Universal Reduction Laws for Macromolecular Protein Dynamics, Allostery, and Kinetics

This paper develops a scalable, thermodynamically reversible stochastic operator for macromolecular dynamics, showing universal reduction laws and improved performance over existing methods.

ExPert infers causal perturbations from cell states across biological and clinical contexts

ExPert infers causal perturbations from cell states by learning prototypes across 12 contexts, achieving high recall with low compute and validating findings in disease models.

Improving real-time search through deep learning-enhanced fragment ion indexing

The paper develops a deep learning-enhanced fragment ion indexing method to reduce memory usage in tandem mass spectrometry searches while maintaining search accuracy and speed.

Wearable Estimation of Knee Joint Contact Forces: SiMPLeSENSE, a Simulated-IMU Pipeline Validated Against In Vivo Loads

The paper develops SiMPLeSENSE, a simulated-IMU pipeline that estimates knee joint contact forces using motion capture and EMG, validated against in vivo measurements with high accuracy.

KH1 Domain of Poly C Binding Protein 1 (PCBP1) Interacts with Utrophin-A mRNA and Modulates Utrophin-A Expression in Mouse Myoblasts

This paper shows PCBP1 interacts with utrophin-A mRNA, represses its expression by promoting nuclear retention, and identifies the KH1 domain as critical for this interaction in mouse myoblasts.

Pathology foundation-model embeddings reveal a morphological landscape of breast lesions beyond diagnostic categories

This paper uses morphology-based foundation-model embeddings to define a morphologic state space across breast lesion categories, revealing a pathobiological continuum associated with tumor development.

Pharmacological activation of mitophagy prevents calcific aortic valve disease by targeting dysregulated mitochondrial-to-nuclear communication

This study shows that activating mitophagy with TJ0113 prevents calcific aortic valve disease by restoring mitochondrial-to-nuclear communication disrupted in diseased valves.

MaxCAT-Net: Hybrid Transformer-Based Skin Lesion Classification and Lightweight Segmentation with Calibration and Cross-Dataset Validation

MaxCAT-Net improves skin lesion classification and segmentation with hybrid architecture and calibration, showing performance on benchmark datasets but with preliminary results.

Cross-species structure-mechanics prediction in the intervertebral disc

This study develops a cross-species framework linking collagen structure to mechanical vulnerability in intervertebral discs using multiscale imaging and modeling.

EigenBrain maps cortical molecular architecture from functional, geometric, and structural eigenmodes

The paper shows functional MRI eigenmodes can approximate specific receptor and transporter distributions in the cortex, while structural and geometric eigenmodes cannot, and applies this to detect molecular differences in autism and schizophrenia.

TSNR: Protein-Specific Signals with Broad Molecular Context for Spatial Protein Prediction

The paper introduces TSNR, a method for spatial protein prediction that uses protein-specific signals and broad molecular context, achieving high correlation on a benchmark.

Uncovering the Multifaceted Role of the Kinetochore Scaffold 1 (KNL1) in Hepatocellular Carcinoma: Integrative Bioinformatics Analysis and CRISPR-Cas9 Experimental Study

This study identifies KNL1 as a key driver of chromosomal instability and immune evasion in HCC through integrative bioinformatics and CRISPR-Cas9 experiments.

QC-espresso: Brewing quality in proteomics

QC-espresso provides an automated quality-control workflow for DIA proteomics data, generating structured reports to assess acquisition quality and quantitative reliability.

Genome-wide CRISPRi-seq maps essential-gene contributions to aminoglycoside susceptibility in Pseudomonas aeruginosa

This study uses CRISPRi-seq to identify essential genes affecting aminoglycoside susceptibility in Pseudomonas aeruginosa, revealing gene-specific fitness effects and potential targets for antibiotic enhancement.

Quantitative imaging of native tissue architecture combined with deep multi-omics reveals the spatial biology of kidney aging

This paper introduces NORIGAMI, a multi-omics pipeline that combines imaging and deep learning to analyze kidney aging, revealing age-related structural and molecular changes.