Hi Federico, Thank you for your interest in our methods. A simple answer to your question is you cannot use log-transformed / negative values in the corrected reference. It has to be at the raw count scale. I have not extensively tested the integration of multiple single cell RNA-seq references. When cell types overlap between multiple references, you can label each batch as a cell state in the, or use the negative binomial regression function provided by cell2location to generate a unified reference matrix. You may also see if there are other (non-linear) single cell data integration methods, such as scVI, that would allow you to project one dataset to the other at the count scale, i.e., non-log transformed / non-latent space. I might be exploring this problem in the future, and will keep you posted. Best, Tinyi