This is the most generalizable work I think I have every done. RNAseq is everywhere these days, and most analysis pipelines have a step for filtering out genes that are expected to be too noisy to interpret. We came up with a strategy to filter based on variance, rather than expression level.

Precision-Based Filtering Facilitates Integration of Conventional and Single-Nucleus Transcriptomes to Identify Time- and Temperature-Sensitive Cell Populations https://www.biorxiv.org/content/10.64898/2026.01.28.702137v1