Gene set or pathway analysis in scRNA-seq data has its own challenges. Since the data is sparse and many genes will be missing for any given cell, some people simply add up all the counts across the gene set. The sum is treated as the expression value of the gene set. I still see this simple approach used in recent papers. Seurat AddModuleScore (and used by CellCycleScoring) is basically the same method, except it additionally subtracts the signal from a control gene set, but that is just shifting the score down.

My concern is that the gene set score can be driven almost entirely by a single gene. If the max value of one gene in a gene set is 100x greater than for others, which certainly happens, then those other ones are not really contributing. What is even the point of using a whole gene set then? Yet, this approach is widely used, so it must be working well enough. What am I missing?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.