Jingyi Jessica Li

The paper introduces a method to quantify diverse perturbation responses and discover novel biological insights in single-cell perturbation datasets.

ClusterDE identifies reliable markers and validates clusters in omics data by controlling false discovery rates through synthetic control, improving accuracy in single-cell, spatial, and bulk transcriptomics.

The paper introduces scDesign3, a statistical simulator that generates realistic single-cell and spatial omics data and infers biologically meaningful parameters for benchmarking computational tools.

scDEED detects unreliable 2D single-cell embeddings and optimizes t-SNE and UMAP hyperparameters using a reliability score across multiple scRNA-seq datasets.

This study evaluates how autoencoder architecture and hyperparameters affect scRNA-seq imputation performance and provides guidelines for optimization.

The paper introduces scGTM, a model for capturing interpretable gene expression trends along pseudotime, showing it can reveal molecular insights in single-cell data.

The paper introduces Markonv, a convolutional layer that models inter-positional correlations using Markov transition matrices, showing effectiveness in analyzing omics data and decoding electrical signals.

The paper re-evaluates differential expression methods on semi-synthetic RNA-seq data without normalization, finding the Wilcoxon rank-sum test still outperforms dearseq in FDR control and power.

scSampler is a fast diversity-preserving subsampling method for single-cell transcriptomic data that outperforms existing methods in speed and accuracy.

The paper shows that popular differential expression methods like DESeq2 and edgeR have high false discovery rates in population-level RNA-seq data, recommending the Wilcoxon rank-sum test instead.

The paper benchmarks PCA against hidden variable inference methods for QTL mapping, finding PCA to be faster, better-performing, and more interpretable.

The paper discusses the controversy around zero-inflation in scRNA-seq data, evaluates impacts of non-biological zeros, and benchmarks different data types.

CAMO evaluates omics congruence between model organisms and humans by analyzing differential analysis, concordance, pathway investigations, and subnetwork detection.

Clipper provides a p-value-free FDR control method for high-throughput data from two conditions, outperforming existing approaches without relying on p-values or distributional assumptions.

mbImpute improves microbiome data analysis by accurately imputing non-biological zeros, enhancing the detection of disease-related taxa in type 2 diabetes and colorectal cancer data.

scDesign2 generates high-fidelity synthetic single-cell gene expression data with gene correlations, enabling benchmarking of experimental protocols and computational methods.

PseudotimeDE identifies differentially expressed genes along cell pseudotime using single-cell RNA-seq data with well-calibrated p-values, outperforming existing methods in false discovery rate control and power.

scPNMF selects informative genes from scRNA-seq data to improve cell type distinction and enable alignment with targeted gene profiling data.

RAD identifies region-associated differentially expressed genes by linking genomic regions to gene expression changes, providing visualization and statistical analysis.

DORGE identifies cancer driver genes by integrating genetic and epigenetic data, finding histone modifications predict TSGs and missense mutations predict OGs.

The paper introduces BiTSC, an algorithm for identifying conserved gene co-clusters across two species using orthology and expression data, showing its accuracy and biological relevance.

An ensemble of six imputation methods, built through the NCI-CPTAC Proteogenomics DREAM Challenge, fills missing values in labeled LC-MS/MS proteomics data, helping most for low-abundance proteins.

The paper introduces CHALM, a methylation quantification method that accounts for cellular heterogeneity, improving the correlation between methylation and gene expression.

The paper introduces a Transformer-based model for annotating Gene Ontology terms in protein sequences, showing improved accuracy over CNN-based methods and incorporating additional features like BLAST motifs and structural data.

This paper evaluates neural network models for embedding Gene Ontology terms and introduces new models based on GCN, ELMo, and BERT, finding that BERT-based embeddings perform best for rare labels.

AIDE improves mRNA isoform discovery by using annotation to enhance precision and reduce errors in RNA-seq data analysis.

The paper quantifies how general mRNA sequence features control translation rates across eukaryotes and identifies key structural and compositional elements involved.

scDesign is a statistical simulator that helps optimize scRNA-seq experimental design by assessing sequencing parameters and generating synthetic datasets for method development.

EpiAlign is a new alignment tool for comparing chromatin state sequences across epigenomic data, identifying common patterns and cell-type-specific features.

The paper introduces two NLP-based methods to measure semantic similarity of Gene Ontology terms using their definitions, achieving better classification accuracy than tree-based methods.