Is it okay to normalize data once more?

Question

[le93jk]

I am building deep learning classification model for scRNA, currently, but datas that I processed different normalization.

My deep learning model use normalize_per_cell in scanpy, but some of datasets have raw expression count data, and some processed RPKM before.

Is it okay to use these datasets together to normalize per cell and train the model?

[biofalconch]

They all need to have the raw data in order to be normalised again together (even perhaps need some integration runs). I don't think performing normalisation on an already normalised dataset is wise, and you might end up drawing wrong conclusions from the data.