A protein chain gains unique properties when it folds into a 3-D configuration. Several serious diseases have been linked to protein misfolding.

One of the grand challenges in computational biology is to be able to predict the three-dimensional structure of a protein molecule from a known sequence of amino acids. Developing new numerical models to simulate the protein-folding process and predict the resulting protein structure is a research goal being pursued by Jeetain Mittal.

Proteins play a central role in the structure and functioning of all living cells. All enzymes are proteins, as are many structural tissues such as muscles and the collagen found in skin, bone and tendons. Proteins are manufactured within the cell cytoplasm on ribosomes, which are bead-like structures associated with messenger RNA. Each polymeric protein chain is generated by linking a large number of amino acids together in a specific order.

The atoms in the polymer chain do not remain stationary but are in constant motion. Their range of motion is restricted by interactions with neighboring atoms and by the amount of free space that is available. Within a few milliseconds, a protein chain folds into a specific 3-D configuration that gives the protein its unique properties. When a protein folds incorrectly, however, things can go wrong. In the past few decades, many diseases, including Alzheimer’s, cystic fibrosis and several cancers, have been linked to protein misfolding.

X-ray crystallography and other investigative techniques have revealed that protein chains fold primarily into two different types of structure: an alpha (α)-helix, in which the chain turns around itself to form a right-handed coil or spiral that is stabilized by hydrogen bonding, and a beta (ß)-sheet, in which the chain folds back on itself.

Mittal and Best have achieved a breakthrough in protein-folding simulation that could shed light on protein-folding mechanisms.

Simulating the folding of an entire protein molecule, which in some cases can contain more than 300 amino acids, is impossible even for today’s most powerful computers. Scientists, therefore, focus their efforts on small globular proteins and smaller segments of longer proteins.

“Many of the current computational models tend to favor the formation of α-helices,” says Mittal, an assistant professor of chemical engineering. “The known propensity of a particular protein sequence to form a specific structure tends to determine the type of model that is used and, therefore, cannot be predictive.”

By simply modifying the part of the force field that represents the torsional stiffness (or twisting behavior) of the chemical bonds in the polymer chain, Mittal, in a collaboration with Robert Best from the University of Cambridge in England, has been able to use the same model, a modified version of the Amber ff03 force field, to predict the structure of four mini-proteins: the GB1 hairpin (containing 16 amino acids) known to form a ß-sheet, the Trp cage (containing 20 amino acids) known to form an α-helix, the Villin HP35 (containing 35 amino acids) known to form α-helices, and the Pin WW domain (containing 37 amino acids) known to form ß-sheets.

“This is the first time that the same model has been used to accurately predict these structures,” says Mittal. The achievement, he adds, represents a major step forward in protein-folding simulation and could lead to a much better understanding of protein-folding mechanisms and why proteins sometimes misfold.

The researchers utilize the computational capabilities of the Biowulf PC/Linux cluster at the National Institutes of Health in Bethesda, Maryland.