To my Knewlodge it's a NO, but there is a ligand-ligand similarity for virtual screening of unkown target structures. There is also a ligand-ligand interaction where scientists studying metal complexes are always studying to to see their biological effect

You could certainly feed the molecules into a standard docking algorithm, but I wouldn't expect much in the way of useful output - those algorithms are optimized for proteins.  You'd probably be better off using something like molecular dynamics and a forcefield designed for small molecules.  Or, if you're just looking at host/guest chemistry, draw the molecules into a standard package, get the guest positioned reasonably in the host, then do an energy minimization to get a better view at the 'final' positioning.

I came across an article (doi: 10.1128/AAC.02347-17), where ligands were docked into a heme extracted from the a protein crystal structure.

Since these ligands are too small for normal protein-ligand docking algorithms, the best approach is running a replica-exchange molecular dynamics simulation with TIGER2h (need less resources then standard T-REMD). For small molecules, 8 replica should be sufficient (300K-450K or 300K-600K, depending on the flexibility). To analyze the simulation, simply calculate the atom-atom distances and perform a principle component analysis.

Article Replica Based Protein Structure Sampling Methods: Compromisi...