columns 24–27, instead of columns 23–26. Pymol displays
them correctly, but VMD does not. This can be corrected by
shifting the residue numbers by one column to the left.- Currently there are over 30 scoring functions, which represent
diverse approaches to the calculation of ligand affinity [30]. It is
a good idea to rescore your results with several different func-
tions and accept ligands as hits only if there is a consensus, i.e.,
when ligands score well with most of the methods. - When preparing ligands with the antechamber utility, the pro-
gram sometimes fails on PDB files. In this case try the mol2
format. However, it is likely that the reasons of these problems
come from erroneous or ambiguous definition of the ligand’s
structure. If the initial structure is not optimized, the missing
CONECT records may be the reason of the failure. - When preparing a complex PDB file with several ligands, make
sure that each ligand has a different chain identifier and also
that you use different residue numbers for different ligands.
Otherwise Amber complains about split residues. - Adding charge neutralizing ions to the system can be done
using the addions command. Another version exists: addions2.
It is longer in execution, but it uses a more sophisticated
algorithm for placement of ions within the periodic box. - When comparing intermolecular energy values calculated with
different versions of the Amber software and different scripts
supplied with it, we obtain results which are not necessarily in
agreement, sometimes not even qualitatively. The most coher-
ent results we have obtained in our calculations come from
Amber 14 (as compared to versions 9 and 12). We compared
the results of energy calculations with the Generalized Born
(GB) and Poisson–Boltzmann (PB) methods by the two most
commonly used scripts supplied with the Amber software:
mm_pbsa.pl written in Perl (PL), and MMPBSA.py, written
in Python (PY). The tests were performed on eight different
receptor-ligand systems (data not published). The results of the
GB method as calculated by the PL and PY scripts are compa-
rable (within10%) for half of the studied systems. In two of
the remaining cases, the PL script gave a result 20–30% higher
than the PY script, in the other two the PL script clearly broke
down, giving unrealistic values. Therefore, we prefer to trust
the results of GB calculations with the PY script.
The calculations using the PB method with the PL script
are consistent with the rest of the results and provide reason-
able values for the cases for which the GB method has not
worked. By way of contrast, the PB method used with the PY
script always gives results which are scaled by a factor of ½ with
respect to all the other values. If we multiply each PB
174 Gre ́gory Menchon et al.