Issues when data matrix contain zero values

[hhageBA]

Hello,
I am using reComBat to remove batch effect from my RNAseq data. My matrix contains zero values for many genes that are not expressed in samples.
This cause an issue of division by zero at the fit_linear_model step 'RuntimeWarning: divide by zero encountered in divide
Z = (data.values.copy() - np.matmul(Covariates[:, num_batches:num_batches+X_covariates_dim], self.beta_x_)'.

When I replace zero values in my matrix by 0.0001 for example, it works without issue.
But its normal in RNAseq matrices to have zero values. Is there an alternative from replacing zero values to not have this error?
Thanks a lot