ff12SB (forcefield Amber 13) is a continuing evolution of the ff99SB force field, primarily developed in the Simmerling group at Stony Brook University.
Several groups had noticed that the older ff94 and ff99 parameter sets did not provide a good energy balance between helical and extended regions of peptide and protein backbones.
Another problem is that many of the ff94 variants had incorrect treatment of glycine backbone parameters.ff99SB improved this behavior, presenting a careful reparametrization of the backbone torsion terms in ff99 and achieves much better balance of four basic secondary structure elements (PP II, b, aL, and aR).
Dihedral term parameters were obtained through fitting the energies of multiple conformations of glycine and alanine tetrapeptides to high-level ab initio QM calculations. We have shownÂ that this force field provides much improved proportions of helical versus extended structures.
In addition, it corrected the glycine sampling and should also perform well for b-turn structures, two things which were especially problematic with most previous Amber force ï¬eld variants.Since 2006, a number of limitations of the ff99SB parameter sets became evident, and a new round of parameter optimization was undertaken. The changes mainly involve torsional parameters for the backbone and side chains. For backbones, ff99SB has been demonstrated to
understabilize helical conformations of transiently folded peptides. Therefore, a principal goal of ff12SB was to predict accurate secondary structure propensities.
A key objective in the ff12SB fitting was to develop parameters that are robust with variation of the local environment,Â including backbone conformation, of which the training set possesses a limited number, and solvent, notably absent from the training. Since side-chain preferences reproducibly vary with
backbone conformation, we employed multiple backbone conformations of each amino acid to partially account for energy backbone-dependence. We also did not preferentially solve our corrections for certain side chain conformers that happen to be stable at a particular backbone conformation of a dipeptide in vacuo.
Where particularly strong non-bonded interactions occur, minor deficiencies in non-bonded models may manifest as significant, structurally-dependent energy errors. This is especially true since Amber charges are not particularly in vacuo charges. Strong non-bonded interactions may also induce strain, exposing errors in bond length or angle representation far away from theÂ ground state. Since the goal is to fit robust parameters describing local dihedral torsion effects
that are appropriate as other structural features may change, we removed from our training any structures where atoms not in a bond, angle, or torsion with each other were particularly close. We also restrained all backbone dihedrals, including hydrogens, to further avoid overly strongÂ vacuum non-bonded interactions.
Together with new corrections for the backbone and the four amino acids addressed in ff99SBildn, this work offers updated side chain dihedral corrections for lysine, arginine, glutamate, glutamine, methionine, serine, threonine, valine, tryptophan, cysteine, phenylalanine, tyrosine, and histidine. ff12SB enhances reproduction of experimentally indicated geometries over ff99SB.
The ff12SB is the current recommended force ï¬eld for proteins and nucleic acids.
Nucleic acid residues in ff12SB use the new (version 3) PDB nomenclature: “DC” is used
for deoxy-cytosine, and “C” for cytosine in RNA, etc. Earlier force fields (which are not recommended!) use “RC” for the RNA version. If you want a single, nucleoside, use “CN”, etc.
leaprc.ff12SB # This will load the files listed below
parm10.dat # ff10 force field parameters
frcmod.ff12SB # ff12SB modifications to parm10.dat
amino12.lib # topologies and charges for amino acids
amino12nt.lib # same, for N-terminal amino acids
amino12ct.lib # same, for C-terminal amino acids
nucleic12.lib # topologies and charges for nucleic acids