We can calculate the distance between residues in a PDB file regarding different parameters like closest atoms, alpha carbon, beta carbon, centroid and etc. Which one of these parameters are better to show physical interaction between residues in a PDB file?
They measure protein-protein (prot-prot) contacts as any Nitrogen, Carbon, or Oxygen element atoms (N, C, O) within 4.0 Angstroms (Å) when measuring the 3d euclidean distance.
If your atomic contacts are not close enough, the amino acids may not actually interact at all, even though they theoretically could. Since in PPIs you generally want to use reliable ground-truth information, it is better to use cautious parameters. Often, in PPIs, the answer to "does it interact?" is not "yes" or "no" but rather "yes" or "don't know / unknown".
You should read my paper, where I did a comprehensive parameter sweep across different thresholds and definitions and benchmarked the resulting empirical potentials on a decoy ranking challenge:
The relevant phrase from the abstract is "compared using 90 different definitions of residue-residue contact". However, it boils down (as always) to the question, "what do you want to do".
Cite me if you use my paper for your research.