Hello,

I have no replicated human cancer cell data, and trying to use edgeR.

In the user's guide "edgeR: differential expression analysis of digital gene expression data",

2.1.1 What to do if you have no replicates

2. Simply pick a reasonable dispersion value.

What of the bcv value is suitable for human cancer cell， 0.4，0.1 or 0.01?

Thanks in advance?

You don't provide a lot of context. But if the data are from a human cancer cell line, generated under tightly controlled conditions in a laboratory environment, you could expect dispersions below 0.01 (i.e., BCVs < 0.1) to be "reasonable". On the other hand, if the data are from actual patients, you can expect BCVs at or above 0.5 - see the Nigerian case study (Section 4.3) in the edgeR user's guide.

Obviously, these are just very general observations. No pre-defined "reasonable" dispersion value is a good substitute for actually having replicates in your experiment.