FLASH-MM speeds single-cell gene analysis for tuberculosis and kidney research

Single-cell RNA sequencing (scRNA-seq) lets scientists compare gene activity inside individual cells, opening up detailed views of how cells differ across conditions. But analyzing these data to find which genes change between groups—called single-cell differential expression—comes with real challenges. Measurements from many cells are not independent: cells from the same person or sample are correlated, and differences between individuals can confound results. On top of that, modern scRNA-seq experiments generate huge matrices of genes by cells, creating steep demands on computing time and memory. To tackle these problems, a team led by corresponding author Gary D. Bader developed a new algorithm called FLASH-MM. FLASH-MM is a fast and scalable approach for fitting linear mixed-effects models (LMMs), a statistical framework that can account for sample correlation and individual variation. The researchers reformulated parts of the LMM estimation procedure specifically for the common gene by cell matrix layout used in scRNA-seq, reducing computational complexity and memory use. That redesign aims to make mixed-model analysis practical on large single-cell datasets without sacrificing the statistical rigor needed to control for structured sources of variation.

FLASH-MM is a new linear mixed-effects model (LMM) estimation algorithm tailored to scRNA-seq data. By refactoring how LMMs are estimated in the context of a gene by cell matrix, the method reduces both computation time and memory requirements. The team tested FLASH-MM through simulation studies using scRNA-seq data and found it to be accurate and computationally efficient. In those simulations FLASH-MM effectively controlled false positive rates—a key concern when testing thousands of genes—and maintained high statistical power to detect true differences in gene expression. Beyond simulations, the authors applied FLASH-MM to real single-cell datasets, including tuberculosis immune and kidney single cell data, and showed that the method accelerates differential expression analysis across these diverse biological contexts. Throughout, the tools and terms mentioned in the work—FLASH-MM, linear mixed-effects model (LMM), and scRNA-seq—are preserved exactly to reflect the method and data used in the study. The results indicate that FLASH-MM can perform rigorous mixed-model analysis faster and at scale on common single-cell data formats.

The practical upshot of FLASH-MM is that researchers can run mixed-model differential expression analyses on larger single-cell datasets without being bottlenecked by computing resources. Because the approach explicitly addresses sample correlation and individual variation, it helps produce more reliable results when many cells come from the same donors or experimental batches. By cutting computational complexity and memory use for the gene by cell matrix layout, FLASH-MM makes it feasible to bring the statistical advantages of LMMs to everyday single-cell studies. That means teams studying immune responses in tuberculosis, researchers analyzing kidney single-cell data, and investigators in other fields can more quickly test hypotheses, control false positives, and retain the power to detect meaningful gene expression changes. In short, FLASH-MM lowers the practical barriers to using rigorous mixed models in single-cell work, enabling faster, scalable differential expression analysis across a wide range of biological questions.