I am using scanpy for cell cycle scoring and regression. According to this tutorial, we should always log-transform and scale data before scoring. Scaling will make the data to be unit variance and zero mean, which will influence the selection of reference genes, so why is this step needed?
The version of scanpy in the tutorial is 0.4.3+7.g2bc93a6, it will need to rescale data after
sc.pp.regress_out. Is this still necessary in later version, such as 1.9.1?