Our initial aims have been to develop the underlying modular framework rather than attempting to embark on a comprehensive reparameterization of lipids. For this reason and given the fact that others have had success using the General Amber Force Field (GAFF)
24 (link) for lipid simulation,
25 (link)–30 (link) we chose to base the initial LIPID11 framework parameter set (excluding the charges) on GAFF. While lipid simulations utilizing GAFF have typically required an applied surface tension to give satisfactory agreements with experiment, GAFF was considered to be a promising starting point for the development of a dedicated phospholipid force field in Amber. Indeed, Gould
et al has recently shown
31 that a simple reparameterization of the Lennard-Jones terms coupled with an automated refinement of key dihedral parameters in the lipid tails can negate the need for an applied surface tension. While GAFF parameters were used for the majority of groups (
Table 1), it was determined that the Glycam force field
52 (link) was more appropriate for the inositol ring of the phosphatidylinositol head group, considering that it is a carbohydrate (
Table 2). Electrostatic and van der Waals interactions are scaled by 2.0 and 1.2 (respectively) in GAFF and the standard Amber protein force fields. In Glycam, 1–4 interactions are not scaled. Accordingly, 1–4 interactions for inositol are scaled by 1.0 while the parameters adapted from GAFF have standard 1–4 scaling for the lipid parameter set. While we have initially employed GAFF van der Waals and bonding parameters we have designed the atom typing assignments of the framework such that future revisions can individually optimize specific parameters without affecting the original GAFF force field. The atom type nomenclature we have developed along with the descriptions of each type are presented in
Tables 1 and
2. To differentiate LIPID11 framework atom types, and thus parameters, from the other Amber force fields, the atom type names consist of a lower case and an upper case character. GAFF, on the other hand, uses all lower case while Glycam and the other Amber force fields use all upper case. The first, lower case letter corresponds to the element the atom type represents, whilst the second character was chosen arbitrarily.
For future flexibility in parameter refinement it was deemed necessary to further differentiate between some of the LIPID11 atom types and those of GAFF. When default GAFF atom types are used in a phospholipid, the same atom type (o) is assigned to ester carbonyl oxygens, phosphodiester sp2 oxygens and to any carboxyl oxygens present (
Table 1). Similarly, the sp3 oxygens of both the phosphate group and the ester linkages are characterized by the oS atom type. In each of the two examples given, the oxygens in the divergent chemical groups might not be equivalent with regards to force field parameters. Thus, these oxygen types have been assigned unique atom types to ultimately allow for different dihedral parameters of the phosphodiester group and the ester linkages connecting the long acyl chains to the glycerol backbone. Such splitting of atom types enables future independent modification of the valence (bond, angle and dihedral) as well as Lennard-Jones parameters for the different chemical environments.
Skjevik Å.A., Madej B.D., Walker R.C, & eigen K.T. (2012). LIPID11: A Modular Framework for Lipid Simulations using Amber. The journal of physical chemistry. B, 116(36), 11124-11136.