Simulating The Folding Of Small Proteins By Use Of The Local Minimum Energy And The Free Solvation Energy Yields Native-Like Structures

Assuming that the protein primary sequence contains all information required to
fold a protein into its native tertiary structure, we propose a new computational
approach to protein folding by distributing the total energy of the
macromolecular system along the torsional axes. We further derive a new
semiempirical equation to calculate the total energy of a macromolecular system
including its free energy of solvation. The energy of solvation makes an
important contribution to the stability of biological structures. The segregation
of hydrophilic and hydrophobic domains is essential for the formation of
micelles, lipid bilayers, and biological membranes, and it is also important for
protein folding. The free energy of solvation consists of two components: one
derived from interactions between the atoms of the protein, and the second
resulting from interactions between the protein and the solvent. The latter
component is expressed as a function of the fractional area of protein atoms
accessible to the solvent. The protein-folding procedure described in this
article consists of two successive steps: a theoretical transition from an ideal
alpha helix to an ideal beta sheet is first imposed on the protein conformation,
in order to calculate an initial secondary structure. The most stable secondary
structure is built from a combination of the lowest energy structures calculated
for each amino acid during this transition. An angular molecular dynamics step is
then applied to this secondary structure. In this computational step, the total
energy of the system consisting of the sum of the torsional energy, the van der
Waals energy, the electrostatic energy, and the solvation energy is minimized.
This process yields 3-D structures of minimal total energy that are considered to
be the most probable native-like structures for the protein. This method
therefore requires no prior hypothesis about either the secondary or the tertiary
structure of the protein and restricts the input of data to its sequence. The
validity of the results is tested by comparing the crystalline and computed
structures of four proteins, i.e., the avian and bovine pancreatic polypeptide
(36 residues each), uteroglobin (70 residues), and the calcium-binding protein
(75 residues); the C alpha-C alpha maps show significant homologies and the
position of secondary structure domains; that of the alpha helices is
particularly close.