You cant. All the samples that you want to compare should be put in the same membrane with good loading charge normalization. Compare 2 different membranes its unreliable and technically almost impossible for the number of variables that are out of your control.

Hi Miguel, thanks for your answer, Even though is pretty different from what I was expecting.

What if I scan the membranes with the same parameters, load the same positive controls in both membranes and finally normalize my samples by the same positive controls?

That would be "OK" for your own record, never for publication. That's not uncommon do that to know more less the levels of high number of samples, in order to select the samples for the results that you want to show in the final publication quality membrane, but inclusive doing that, you could have some surprises (e. g. the loading normalization is not correct and the level of expression of your protein of interest change)

Ok Miguel, thanks for the precious advice. Best Regards

Sure, you can. A normalization should be don in one Western blot leading to absolute values after semi-quantification (visually or digital via any tool (e.g. Gel-Pro Analyzer)). Calculating the ratio compared to any control should be done (which should apply in all Western blots which have to be compared).

Finally, these ratios can be used to compare data from different Western blots. Results are presented as "relative expression".

This method is widely accepted and easy to perform (e.g. see Cao C et al., Cancer Manag Res. 2019 Nov 14;11:9675-9683. doi: 10.2147/CMAR.S213830.

or

Ly LD et al., Mol Cells. 2020 Jan 13. doi: 10.14348/molcells.2019.0223.

and so on...).