Dongyuan Song, PhD

Our primary research interest is to develop computational methods for analyzing modern “omics” data, especially for single-cell and spatial transcriptomics. Our research combines statistical modeling, bioinformatics, and machine learning to provide a more rigorous interpretation of biological signals. Some specific topics include: 1. Generation of realistic in silico single-cell and spatial multi-omics data; 2. Computational modeling of gene co-expression in single-cell and spatial transcriptomics; 3. Detection of differential expression in single-cell and spatial transcriptomics; 4. Gene selection, dimensionality reduction, and cell subsampling for large-scale datasets.

[2025] PseudotimeDE-fast: fast testing of differential gene expression along cell pseudotime. [2025] ClusterDE: A Statistical Software Package for Removing Double-Dipping Bias in Post-Clustering Differential Expression Analysis. [2025] Systematic benchmarking of computational methods to identify spatially variable genes. [2025] Worm Perturb-Seq: massively parallel whole-animal RNAi and RNA-seq. [2025] Decoding heterogeneous single-cell perturbation responses. [2024] DNA binding analysis of rare variants in homeodomains reveals homeodomain specificity-determining residues. [2024] scDesign3 generates realistic in silico data for multimodal single-cell and spatial omics. [2023] scReadSim: a single-cell RNA-seq and ATAC-seq read simulator.