deep learning

Topology-dependent FRET efficiency in living cells via N-C swapping of fluorescent protein fusions

This study shows that the topology of fluorescent protein fusions, not just distance, affects FRET efficiency in living cells, with donor-acceptor configurations showing higher efficiency than acceptor-donor.

Toward a Transferable Representation of Protein Dynamics for Functional Prediction and Mechanistic Discovery

This paper introduces TIDGN, a deep learning method to encode protein dynamics into a transferable latent representation, achieving high AUROC in function prediction across different proteins.

Slice, Subtract, Segment: Efficient 2D and Physiologically Informed 3D Brain Tumor Segmentation

This paper evaluates two approaches for improving brain tumor segmentation: a novel 2D U-Net architecture and a multi-channel input representation using cross-modal subtraction maps, showing both contribute to performance improvements.

ASTRA enables zero-shot super-resolution of spatial transcriptomics across capture geometries

ASTRA enables zero-shot super-resolution of spatial transcriptomics across different capture geometries, improving accuracy and efficiency compared to existing methods.

Spatial Transcriptomics As Rasterized Image Tensors (STARIT) characterizes cell states with subcellular molecular heterogeneity

STARIT converts spatial transcriptomics data into image tensors for deep learning analysis, enabling subcellular heterogeneity-based cell state identification.

Deep Learning-Meets-Active Surfaces: A Unified Framework for Multi-Object Segmentation and Shape Analysis in 3D Fluorescence Microscopy

This paper introduces Nagini3D, a 3D image segmentation method combining deep learning and active surfaces for improved shape analysis and consistency across 3D volumes.

Intrinsic dimensionality of deep learning representations reveals cell death-associated heterogeneity in Parkinsons disease iPSC-derived neurons

The paper uses deep learning to identify cell death-related heterogeneity in Parkinson's disease iPSC-derived neurons by analyzing intrinsic dimensionality.

pre-miRBench: a multispecies benchmark and reference model for animal precursor microRNA prediction

The paper introduces pre-miRBench, a multispecies benchmark for pre-miRNA prediction, showing its model outperforms existing predictors in cross-species and cross-family tests.

Joint protein, mRNA and DNA sequence design and optimization with nucleotide-level Potts models

A structure-conditioned nucleotide-level Potts model designs protein and coding sequences jointly, enabling tRNA adaptation and RNA motif embedding beyond synonymous codons with designability comparable to Caliby.

BettaAI: a machine-learning object detection model for the quantification of aggressive displays in the Siamese fighting fish Betta splendens

BettaAI is a deep learning model that accurately detects and quantifies aggressive displays in Siamese fighting fish using video analysis.

RGI-Toolkit: Differentiable Restraints for Controllable Biomolecular Structure Prediction

RGI-Toolkit enables controllable biomolecular structure prediction by integrating experimental restraints into diffusion models without retraining.

A Comparison of Mechanisms Driving Lesion Outcomes during Lung Tumor and Tuberculosis Granuloma Formation

This paper compares the mechanisms driving lesion outcomes in lung tumors and tuberculosis granulomas using agent-based models to study immune-tumor interactions and treatment responses.

PIE: Generalizing perturbation effects across unseen perturbations, contexts and datasets

Predicting differentially expressed genes and expression shifts per condition from database embeddings plus per-gene evidence, PIE reaches 1.2–3.2 times the best baseline's AUPRC on unseen contexts, perturbations and datasets.

Selective suppression and biasing of chemokine receptors CCR9 and ACKR4 through targeting CCL25 with de novo miniproteins

The paper designs miniproteins that selectively block or bias chemokine receptor signaling by targeting CCL25, demonstrating their ability to suppress CCR9 and modulate ACKR4 responses.

golgi: open-source software for automated nerve model generation and recruitment simulation

golgi automates nerve model generation and recruitment simulation from images, offering a scalable open-source platform with Python API and GPU acceleration.

Agentic campaign control for high-throughput de novo binder design

This paper introduces T-REX, an agentic framework for protein binder design that adaptively allocates compute resources across methods, achieving higher throughput of structurally distinct hits than baselines.

A Structural Model of Transcriptional Activation by Mediator

The paper models human transcription factor activation domain-Tail module interactions of Mediator using AlphaFold 3 to understand how these interactions influence the Mediator-Pol II complex structure and transcriptional activation.

A Graph-Attentive GAN for Rare-Cell-Aware Single-Cell RNA-Seq Data Generation

GARAGE generates realistic synthetic scRNA-seq data by addressing high-dimensional, small-sample and rare-cell challenges using graph attention and adversarial generation.

Functional Deorphanization of Tumor-reactive TCRs Uncovers a Landscape of Shared, Immune Dominant Antigens across Cancers

This study introduces TWISTAR, a platform to identify tumor-reactive TCRs and their antigens, revealing shared immunogenic targets across cancers and immune dominance.

Answering clinicians' questions over trial evidence tables with verifiable, feedback-driven language models

FD-SCoPE answers clinician questions about clinical trial evidence tables using language models, providing verifiable answers and learning from expert feedback.

Spatial Sampling and Temporal Dynamics Shape Motor Decoding from Broadband Intracortical Signals

The paper compares neural signal features from different electrode configurations and decoding strategies to identify factors affecting motor decoding performance.

LGTM: Gaussian Process Modulated Neural Topic Modeling for Longitudinal Microbiome

LGTM couples a neural topic model with additive basis-function Gaussian processes, learning microbial co-abundance topics and how covariates relate to their trajectories in longitudinal gut microbiome data.

AI-assisted quantification of the clonogenic assay

This paper develops an AI framework for automated clonogenic assay analysis, showing it matches expert performance and enables continuous phenotypic profiling of radioresistant cells.

Behavioral characterization of dynamic facial expression perception in rhesus monkeys (Macaca mulatta) using naturalistic and synthetic stimuli

This study shows rhesus monkeys can categorize facial expressions and generalize to novel stimuli, with performance influenced by expression components and signaler characteristics.

Robust prediction of drug combination side effects in realistic settings

DCSE predicts drug combination side effects using a deep learning model, outperforming existing methods in realistic settings with limited data.

Slide-level batch structure limits histology-guided supervision of transcriptomic foundation models

This paper investigates whether histology can improve transcriptomic foundation models by providing cross-modal supervision, but finds limited benefits due to slide-specific structure in gene-expression embeddings.

PRIZM: Combining Low-N Data and Zero-shot Models to Design Enhanced Protein Variants

PRIZM uses low-N data and zero-shot models to rank protein variants, improving design efficiency with minimal experimental data.

Ref(2)P/p62 protein pleomorphism confers tissue-specific function

One Drosophila Ref(2)P polypeptide takes different structures in midgut and muscle, driving autophagy or mitochondria-associated degradation; deleting disordered region IDR2 switches muscle Ref(2)P to midgut state.

trFlow: ultrafast all-atom protein conformational ensemble generation via conditional flow matching

trFlow generates all-atom protein conformations faster and more accurately than existing methods by using flow matching with three key innovations.

CellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics

CellART extracts single-cell information from high-resolution spatial transcriptomics data across multiple platforms, demonstrating efficiency and accuracy in cancer datasets.