Science in Society Archive

Water Smoothing Protein Relationships

How proteins get into shape through water chains. Dr. Mae-Wan Ho

Proteins consist of linear chains of amino acids joined up together when first synthesized in the cell, but soon adopt various secondary and tertiary structures. Most of them end up in a folded shape that is important for the function they serve; at least this is the conventional view (see “Strong medicine for cell biology”, SiS 24).

For a long time, researchers have acknowledged the importance of water in protein folding, but only in a negative role, in that it is the avoidance of water in hydrophobic (water-hating) interactions that determines how the chain folds up. As yet, most computer algorithms for predicting the folded structure of proteins do not take water into account.

When the protein chain folds up, amino acids from widely separated parts of the chain come into contact, but it is difficult to predict which ones will end up in contact. The team led by Peter Wolynes in UC San Diego and University of Illinois at Urbana in the United States first investigated what would happen if these specific interactions between amino acids were mediated by water. They developed a model based on information available in the bioinformatic database to derive potentials both for direct contact between pairs of amino acids on the protein chain and indirect contact through one or more molecules of water.

When both potentials were used in computer simulations, a smooth recognition of the diverse binding partners takes place, in contrast to runs in which only direct contact potentials were used. 

The researchers then went further, and showed that water plays an important role not only in helping amino acids find their contact partners, but also in how the whole chain folds up.

For their computer simulation, they used the ‘AM Hamiltonian’ molecular dynamics model as a starting point. A ‘Hamiltonian’ is an energy function - named after the mathematician who invented it - which is characteristic of a given process. The AM Hamiltonian has two main components: one based on the physics of the backbone of the protein chain, the “backbone” component, and the other based on knowledge of the specific amino-acid sequence in the chain and their energy potentials, collectively referred to as the “AM/C (contact)” component of the Hamiltonian.  The ‘AM’ part describes interactions between all pairs of amino acids separated in sequence between 3 and 12 neighbours, which have been obtained for a set of 156 proteins in the bioinformatic database. The ‘C’ part applies to contacts between amino acids separated by >12 neighbours in the sequence, and consists of several potentials applying to different range of contact distances between the amino acids. 

Wolynes’ group incorporated into their model the first direct contact C potential, and added a ‘water potential’ W for contact distances between 6.5 and 9.5Å (an angstrom, or 10-10m).

This gave a significant improvement in protein structure prediction. “Wetting” the Hamiltonian with water improved the predicted structure, especially for large proteins.

It seems that highly charged amino acids don’t like to be in direct contact, and such contacts are unstable, whereas the contact is stabilized if mediated by water. That is because a lot of energy has to be spent getting these hydrophilic (water-loving) groups to let go of water molecules that they are already bound to, so they tend to contact other highly charged groups via one or two water intermediates. Even more interesting is that not only oppositely charged residues attract each other through water, but so do groups with the same charge, which suggests that the one or the other of the groups in contact must be changing its sign, or else the contact partners fluctuate coherently together, ie, take turns being the charged or uncharged (rather like dancing rapidly back and forth perfectly in step).

Simulations were done on 14 proteins, from small to large. For all but three proteins, including the water potential significantly improved the fit to the native folded protein; some of the improvements were very striking, and improvements were especially consistent for proteins larger than 115 residues. 

The results show that water not only induced protein folding and binding, but also actively participates via long-range water-mediated contacts. Adding water may improve protein docking, protein and drug design strategies, and contribute to understanding the important role played by water in the function of proteins.

There are other simulations of protein folding in water that provide good quantitative agreement with experimental data. For example, the research team at Stanford University has a model that takes explicit atomistic account of all water-water, water-protein interactions in order to arrive at the folded state [3]. A word of warning is in order in case you think that this computer simulation accurately represents how proteins fold. It takes about 300 years for a computer to simulate a small peptide of 23 amino acid residues to fold into shape in 3938 water molecules. By running simulations simultaneously on some 140 000 individual computers around the world, the researchers took just over three weeks [4]. The protein itself, however, folds to perfection in several microseconds.

Article first published 25/10/05


  1. Papoian GA, Ulander J, Eastwood MP, Luthey-Schulten Z, and Wolynes PG. Water in protein structure predictions. PNAS, 2004, 101, 3352-7.
  2. Levy Y and Onuchic JN. Water and proteins: a love-hate relationship. PNAS 2004, 101, 3325-6.
  3. Rhee YM, Sorin EJ, Jayachandran G, Lindahl E, Pande VS. Simulaions of the role of water in the protein-folding mechanism. PNAS 2004, 101, 6456-61.
  4. “New approach uses computers to study how water molecules affect key biological function”, Joy Ku, Stanford Report 28 April 2004,

Got something to say about this page? Comment

Comment on this article

Comments may be published. All comments are moderated. Name and email details are required.

Email address:
Your comments:
Anti spam question:
How many legs on a tripod?