Denaturation does not change the molecular weight of a protein, unless there is a release of a noncovalently-bound cofactor. The fluorophore of GFP is formed from its amino acids and is covalently bound.

The apparent molecular weight based on experimental observations such as analytical ultracentrifugation, size exclusion chromatography, electrophoresis or dynamic light scattering may differ from the actual molecular weight because of the effect of the shape of the protein on its hydrodynamic properties. A denatured protein has a more extended shape than the folded structure, and therefore will have different hydrodynamic properties. Perhaps that has something to do with the calculation you saw.

Thank you, Adam. I agree with you that the molecular weight of a protein should be determined by its amino acid sequence. It really puzzled me where the computational formulas come from. Anyway, thanks again!

It seems that someone developed a regression model for protein molecular weight in which molecular weight is 10^(-1.57x+5.38), where x is the measurement. My guess is that x is the relative mobility (values from 0 to 1) of the protein in some separation system such as electrophoresis.