Single-cell and spatial transcriptomics

GPSNorm fits slide, spatial technical and biological effects in one Bayesian negative binomial model, carrying normalization uncertainty into differential expression and lowering log-fold change error in simulations.

PGD improves visualization of single-cell trajectories by smoothing cell-level features along pseudotime, enhancing continuity and structure in complex biological processes.

The paper evaluates single-cell perturbation prediction models by developing a taxonomy and tool to standardize benchmarking protocols across datasets.

Spatium is a foundation model for spatial proteomics that learns stable cell representations from constrained protein measurements, enabling accurate cell identity recovery and spatial microenvironment analysis.

The paper develops a cross-species deep learning framework to annotate and attribute ageing states in haematopoietic stem and immune cells, achieving high AUROC scores in human data.

scEvolver enables continual cell-type annotation across expanding single-cell atlases, revealing cellular state transitions in disease contexts.

The paper uses AI to model CD8 T-cell responses during ICI therapy, identifying a latent effector state that conventional methods miss.

The paper introduces SFinx, a framework for spatial functional analysis of transcriptomic data, revealing spatially varying pathway interactions in lupus nephritis.

This paper benchmarks differential expression classification in single-cell perturbation models, finding that simple baselines often outperform deep learning models.

The paper constructs a multi-omics single-cell atlas of human RPE and choroid, identifying cell types, aging patterns, and genetic risk factors for AMD.

GatorPrism integrates spatial multi-omics data by balancing cross-omics consensus with modality-specific signals, achieving strong performance across clustering metrics and revealing tissue structure organization.

The paper uses deconvolution of bulk RNA-seq to generate cell-type-specific targets for predicting gene expression from personal genome sequences, finding that sequence-derived models outperform genotype-feature baselines.

Autocorrelation-aware bivariate Moran's I, bivariate Gaussian process tests and generalized additive models (R package sbivar) test cross-modality colocalisation on disjoint coordinates; many existing permutation tests inflate false positives.

scSAID provides a cross-species single-cell skin atlas to study psoriasis, revealing species-specific molecular signatures and identifying a novel drug target.

This paper introduces CurvSeq and CurvSee to analyze prostate cancer morphology and molecular profiles, identifying cancer-associated genes and morphomechanical niches linked to progression.

The paper provides a high-resolution molecular and cellular analysis of Broca's area and motor cortex across developmental stages, identifying distinct cellular signatures and vulnerabilities to neurodevelopmental disorders.

This paper introduces a lysis-independent method for joint host-pathogen single-cell profiling using bacterial surface display, enabling analysis of both host and pathogen features in infected cells.

SpatialJEPA transfers spatial context from spatial multiomics data to dissociated non-spatial data through graph-context distillation, improving alignment and recovery of spatially organized programs in mouse brain multiomics.

This paper introduces SLMC to select minimal gene programs for disease-aligned representations in single-cell RNA-seq data, finding greedy methods outperform LASSO with small gene budgets.

IntraTalker+CrossTalkeR models intracellular and intercellular signaling in single-cell data to uncover cell-state changes and novel signaling roles.

This paper introduces Phenoverse, a deep learning framework that learns disease state representations from single-cell transcriptomic data, enabling disease prediction and severity assessment using binary labels.

The paper develops a method to analyze high-coverage single-cell RNA-seq data to identify gene expression patterns in C. elegans cells without dimensional reduction.

The paper provides CITE-seq data of bone marrow cells from septic and sham mice to study transcriptional and surface marker changes in sepsis.

Easy-Multiome enables joint profiling of gene expression and chromatin accessibility in plant single cells with high efficiency and minimal workflow changes.

Pan-human Azimuth maps diverse human cells into a unified hierarchical typology, enabling cross-tissue comparisons and spatial transcriptomic annotation.

scRepresenter provides a workflow to compute, integrate, and benchmark cellular representations from scRNA-seq data using different methods for complex disease analysis.

The paper introduces multiScaleAC, a method to analyze cell-cell interactions across spatial scales using a modified Moran's I with varying kernel bandwidths, showing improved statistical power and reduced bias.

FloREN improves patient-level sample stratification and biomarker discovery by modeling single-cell data as an interpretable graph transformer network for immune regulatory analysis.

The paper develops MAESTRO, a self-supervised deep learning framework to transform immune profiles into embeddings capturing immune health, enabling clinical prediction and precision immunotherapy.