Yuanhua Huang

BLOOM generates developmental cell-expression distributions across time using conditional flow matching, outperforming baselines in gene and cell-expression correlations.

CLEAR-ST corrects mRNA lateral diffusion in spatial transcriptomics by modeling contamination with a physics-informed probabilistic framework, improving spatial domain recovery and downstream analysis accuracy.

stCNASim is a spatial RNA-seq simulator that benchmarks copy number inference methods in spatial transcriptomics.

SpatialFormer integrates spatial and gene expression data to improve tasks like batch correction and cell-type annotation in spatial transcriptomics.

The paper identifies cell-state-specific mechanisms of 5-azacytidine response in myelodysplastic neoplasms, showing distinct differentiation and transcriptional changes in cycling progenitors and quiescent stem cells.

SHINE integrates spatial transcriptomics and metabolomics data to identify microenvironments, outperforming existing methods in segmentation and biomarker co-localization.

BRIDGE generates virtual spatial transcriptomes from histology images, improving cancer diagnosis accuracy in few-shot settings and outperforming existing models.

CRAK-Velo improves RNA velocity estimation by integrating chromatin accessibility data, enabling accurate cell-type deconvolution and insights into gene-chromatin interactions.

This study uses single-cell multiomics to show that AZA response in MDS involves hematopoietic regeneration from mutated HSPCs and that AZA-refractive sub-populations contribute to disease progression.

Hist2Cell predicts fine-grained transcriptional cell types from histology images with high accuracy, enabling cellular mapping and cancer prognosis.

SparseAEH accelerates spatial transcriptomics analysis by using sparse Gaussian kernels, achieving faster computation with accuracy comparable to existing methods.

This study uses mitochondrial variants as barcodes to track clonal dynamics and spatial organization in human hematopoietic organoids.

RoBep predicts B-cell epitopes by modeling spatial continuity, improving performance on benchmark datasets and providing antibody-antigen binding regions.

Pseudodynamics+ reconstructs population dynamics from time-resolved single-cell data using physics-informed neural networks, revealing dynamic changes in cell state parameters.

The paper identifies hypoblasts as the origin of the first blood cells in human and non-human primate embryos, showing they contribute to hemoglobin+ phagocyte-like cells and sustain oxygen levels before heart formation.

scTail detects polyadenylation sites using reads 1 of 3' scRNA-seq and quantifies their usage with reads 2, revealing differential alternative PAS usage in biological processes.

This study evaluates genomic language models' ability to predict personal gene expression using paired genome-transcriptome data, finding limited success in capturing inter-person variability but improved performance for seen genes in new individuals.

AutoPM3 automates PM3 evidence extraction from scientific literature using LLMs, improving variant interpretation for rare diseases.

The paper identifies a TNF-α responsive cryptic promoter driving ERVK-7 overexpression in LUAD, revealing cell-type and inflammatory signal-specific regulation.

This paper introduces inDecay, a few-shot learning system for predicting CRISPR editing outcomes in cell lines and zygotes, achieving high accuracy with minimal data.

StaVia integrates spatial and temporal data with higher-order random walks to improve cell lineage analysis and visualization in developmental atlases.

XClone detects allele-specific copy number alterations from scRNA-seq data by smoothing signals on cell and gene coordinate graphs, enabling robust identification of haplotype-aware CNAs and subclones.

This paper introduces TransImp, a model for imputing missing features in spatial transcriptomics with uncertainty estimation and spatial regularization, improving reliability in downstream analyses.

CellContrast reconstructs spatial relationships in single-cell RNA-seq data using contrastive learning, outperforming existing methods and enabling accurate spatial analysis.

CLADES estimates clone-specific kinetics and fate bias in human cord blood LT-scSeq data using a NeuralODE-Gillespie hybrid approach, improving meta-clone fate prediction accuracy.

SNPmanifold uses a variational autoencoder to identify clonality and lineages in single cells, outperforming existing methods in complex datasets.

MetaQuad identifies shared polymorphic variants in metagenomic samples using density-based clustering, reducing false positives while maintaining true-positive rates.

NeuroVelo reconstructs cellular dynamics and identifies gene regulatory networks from single-cell transcriptomic data using dynamical systems theory.