Luyi Tian

MINT infers cell-resolved spatial transcriptomes from paired histology and Slide-tags data, improving alignment and expression recovery in lung adenocarcinoma.

The paper benchmarks 13 models for predicting single-cell perturbation responses across 25 datasets, finding performance varies with perturbation effect size and evaluation metric.

This study uses spatial transcriptomics and lineage tracing to show how tumor subclones drive microenvironmental changes that promote metastasis in a mouse model.

The paper introduces a framework for self-auditing AI in drug discovery, using verifiers to validate predictions and adjust objectives based on experimental results.

The paper develops DynaALS, a disease dynamics atlas for ALS that maps patient states and molecular programs using longitudinal medical records and multi-omics data.

RVQ-Alpha adapts large language models for single-cell analysis by integrating hierarchical tokenization and reinforcement learning, achieving performance comparable to task-specific models on a multi-task benchmark.

Garfield integrates single-cell and spatial multi-omics data into a shared space, outperforming existing methods in identifying spatial niches and revealing immune microenvironment heterogeneity.

This paper introduces SuperSCC, a method for discovering cell identities and gene modules across datasets without integration, showing improved annotation accuracy and conserved immune cell status identification in lung microenvironments.

FLAMESv2 is a flexible R/Bioconductor package for analyzing long-read single-cell RNA-seq data, demonstrating field-leading performance in isoform analysis and cell-type identification.

AtlasAgent uses a vision-language model to rapidly evaluate large-scale single-cell data integration, outperforming traditional methods in speed and efficiency.

OriGene identifies therapeutic targets by integrating diverse data modalities and outperforms human experts and models in accuracy and robustness.

stPipe provides a flexible R/Bioconductor pipeline for preprocessing spatial transcriptomics data from multiple platforms, enabling standardized analysis and benchmarking.

PertFormer, a multimodal foundation model, predicts zero-shot functional perturbations and cell fate dynamics, outperforming existing methods and identifying novel tumor treatment targets.

This study compares six sequencing-based spatial transcriptomic methods using reference tissues, highlighting molecular diffusion as a key variable affecting resolution and identifying unique attributes of spatial transcriptomic data.

The paper introduces scRaCH-seq, a method for single-cell long-read sequencing that improves detection of isoform usage and mutations in targeted genes, demonstrating its application in CLL samples.

Slide-tags enables spatial localization of single nuclei for multi-modal spatial genomics, achieving high-resolution mapping and multiomic profiling in various tissues.

The paper benchmarks long-read RNA-sequencing tools using synthetic data to evaluate isoform detection and differential expression analysis performance.

The paper develops FLT-seq and FLAMES to characterize full-length isoforms in single cells of human and mouse, identifying thousands of unannotated isoforms and linking them to functional modules and drug resistance mutations.

This study benchmarks 9 scRNA-seq preprocessing workflows, finding that while they vary in gene detection and quantification, downstream analysis with effective normalization and clustering methods leads to consistent clustering results.

This study analyzes nanopore long-read RNA-seq data using short-read tools to assess differential gene and isoform expression in synthetic and mouse datasets.

The paper benchmarks scRNA-seq analysis methods using synthetic datasets generated from mixed cancer cell lines, identifying methods that perform well across tasks like normalization, clustering, and integration.

SIS-seq links gene expression with cell fate by analyzing clonal progeny, identifying early gene signatures of dendritic cell subset development.

This paper compares clustering methods for 10X scRNA-seq data using human datasets, finding that clustering results vary widely and are influenced by ribosomal gene expression.

scPipe provides a flexible R/Bioconductor pipeline for preprocessing scRNA-seq data, handling multiple barcoding strategies and generating quality reports for downstream analysis.