Recently, a highly efficient and flexible non-parametric method for detecting gene expression patterns across multiple time points has been developed by Professor Shiquan Sun's team at the School of Public Health, Xi'an Jiaotong University. This method, called TDEseq, short for Time-Dynamic Differential Expression genes in time-series single-cell RNA sequencing data, has recently been published in Genome Biology.
TDEseq utilizes a Linearly Additive Mixed Model (LAMM) to fit the relationship between individual gene expression values and time points. It detects genes with specific expression patterns in the dynamic time series of gene expression levels by introducing spline functions with shape constraints to represent the dynamic changes, introduces random effect terms to control heterogeneity between samples, and ultimately generates statistically rigorous p-values. TDEseq not only ensures higher testing power but also achieves better control over the false discovery rate, especially when dealing with heterogeneous multi-sample scRNA-seq data.
TDEseq demonstrates excellent performance in time-series scRNA-seq data such as cancer cell line drug response, mouse liver embryonic development, lung adenocarcinoma progression, and NK cell response to SARS-CoV-2 virus infection. Taking mouse liver embryonic development as an example, TDEseq identifies 20% more dynamic differentially expressed genes than tradeSeq. The dynamic differentially expressed genes identified by TDEseq not only have clear temporal dynamic expression patterns but also exhibit strong correlation with the liver embryonic development process.
Related paper information: https://doi.org/10.1186/s13059-024-03237-3
