I am wondering what is the % or number of Aa different to consider a protein as a new type or new variant?. I am working with new proteins (Kazal-type protease inhibitors). If you can suggest a reference would be great!
I'm not aware about any specific % or number of AA different that might be a definition for a new protein. I would suggest to consider any protein as a new protein if the revealed structural differences lead to a new function of the protein (excluding any post-translational modifications). In your case, if your protein had a different protein inhibitor specificity (or different binding sites, or different kinetics of binding) you might consider this protein as a new protein.
In the case of enzyme, a slight change of the catalytic residue can change the activity by more than 85% percent. In my bacterial G protein from Klebsiella pneumoniae, (KpNFeoB) a point mutation of the catalytic residue Thr37 to Ser37 (NFeoB-T37S) only has ~13% activity compared with its WT.
In the case of (protein) inhibitors to a protease, you probably want to look at protein-protein interactions (PPIs) according to the solved structure(s). In PPI, binding surface may not be of the same importance. Namely, the binding energy will not evenly distribute across the whole PPI binding surface. Thus, the mutations in the more important regions (that contribute to more binding energy) will make the mutant protein a new species. If you do have a structure, it will be easier to classify the residues. If you don't have a structure, then sequence alignments using bioinformatics might be a reasonable approach to predict these key residues.