Other "best" routes from Point A to Point B aren't so simple. Ask anyone who has ever meandered helplessly through a myriad of small towns on a shortcut to Chicago.
How, then, does one find the shortest, cheapest, quickest way to do anything, from travel cross country to assemble a widgit? Especially if there are many pretty-good solutions?
Often, the answer lies in statistical physics. Since the 1980s, scientists have used a mathematical technique called "simulated annealing."
"The process mimics a slow, cooling process in which the search is gently coaxed toward the best solution," said Dr. Ulrich H. E. Hansmann, an associate professor of physics at Michigan Technological University said. "The method has been used very successfully over the years, but it can be notoriously slow and sometimes needs operator intervention."
Recent work by Hansmann and his colleague Luc T. Wille, of Florida Atlantic University, which appears in Physical Review Letters, Vol. 8, Issue 66, promises to overcome these drawbacks. Hansmann and Wille have designed an "Energy Landscape Paving" method which circumvents the problems of the annealing algorithm.
Their new method is fast and automatic, so much so that they use it to address the notorious protein folding problem, which has baffled scientists for decades.
Proteins are the stuff that we are made of. They are long strings of amino acids, and determining the amino-acid sequence of proteins isn't too difficult. But as proteins are formed, they almost immediately begin folding into incredibly complex structures that interact with other proteins, often in lock-and-key type arrangements. So far, scientists have not been able to determine or predict what form most proteins actually take.
Without understanding proteins' 3D nature, we can't understand what they do or how they do it, so we can't truly understand how organisms, from yeast to human beings, function in this world.
Hansmann and Wille have brought us one step closer, however.
Their new method has found the folded configuration of two test molecules. They stress that their technique is no panacea; for large proteins, the computing time may still be prohibitive. But for smaller biomolecules, it promises better results, faster.
"If you know the structure, you can predict the function," Hansmann said. "This could be important for the development of new chemicals and materials, including pharmaceuticals."
Journal
Physical Review Letters