Most molecules are free to assume a variety of conformations by rotating about single bonds, and which conformations they prefer can have a great influence on their properties. For example, enzymes are active as catalysts and subject to biochemical controls on their activity when the polypeptide chain is correctly folded in space (the native state) and inactive when incorrectly folded. Conformational analysis has been very successful in treating molecules with few degrees of freedom by approximating the free energy as a function of conformation, and then locating regions of conformation space having relatively low energy. For molecules as large or larger than small peptide hormones, however, there are an astronomical number of local energy minima scattered throughout a conformation space of very high dimensionality, and only a vanishingly small fraction of these have low enough energy to be physically significant. A thorough search would require an amount of computer time that increases exponentially with the size of the molecule such that a decapeptide is well beyond the reach of any foreseeable computers. It does us little good to sequence the entire genome of a virus (or eventually the human genome) if we are unable to predict the folding of the corresponding proteins and hence their function. Similarly genetic engineering needs to know what alterations will improve a protein's properties, such as increasing its thermal stability or changing an enzyme's specificity. Energy embedding is a technique we have pioneered for sidestepping this problem entirely by treating the molecule in the computer as if it existed in many more than three dimensions. Our long term goal is to apply energy embedding to the prediction of the low-resolution global folding of proteins. We are learning that successful predictions are guided entirely by a potential function that may have numerous local minima, but must prefer the native conformation in a global sense. Thus developing a suitable potential is our top priority, and we have invented a systematic method for carrying this out, based on linear programming. Since most tests of molecular mechanics potential functions examine their properties only in the neighborhood of experimentally determined conformations energy embedding Is a unique tool for validating their global character. Therefore another short term goal is to examine their properties only in the neighborhood of experimentally determined conformations, energy embedding is a unique tool for validating their global character. Therefore another short term goal is to examine the global predictive ability of standard potential functions, such as AMBER and MM2, on small molecules. A third immediate task is to vectorize our computer programs in order to make larger molecules feasible subjects of study.
Showing the most recent 10 out of 15 publications
