ADL: Different unbound energy with pre-compiled autodock4 executable

Michele Lunelli efunit at
Fri Mar 28 09:47:28 PDT 2008

Dear all,

I am docking a small peptide on the surface of a protein, using autodock 4.0.1 
compiled on my machine (Suse linux 10.1) with g++ from the source files.
For comparison I tried to run the same input dpf file using the pre-compiled 
version downloaded from the autodock site 
(autodocksuite-4.0.1-i86Linux2.tar.gz), taking care of using the same seeds for 
the random number generator. I was quite surprised to see that the results are 

Analyzing the log files, I found that the reason of the differing results is in 
the calculation of the energy of the unbound "extended" state. In the first case 
(autodock compiled by me) everything seems to be fine: the ligand (that was 
already minimized) has an internal energy of about -1 kcal/mol. In the second 
case (autodock pre-compiled) there is one couple of atoms (out of 5339 couples) 
that has a very high vdW+Hb energy, about +30 kcal/mol, because the distance is 
reported to be only 2.19 A (but in the initial model of the ligand the distance 
is actually 3.9 A!). Consequently, the energy of the "extended" state of the 
ligand is too high, and autodock searches for a lower energy unbound state.

Is this a bug of the pre-compiled version? Am I correct in considering the first 
case as the correct one?

Thank you in advance,
Michele Lunelli

More information about the autodock mailing list