Variable selection for QSAR by artificial ant colony systems.

Derivation of quantitative structure-activity relationships (QSAR) usually involves computational models that relate a set of input variables describing the structural properties of the molecules for which the activity has been measured to the output variable representing activity. Many of the input variables may be correlated, and it is therefore often desirable to select an optimal subset of the input variables that results in the most predictive model. In this paper we describe an optimization technique for variable selection based on artificial ant colony systems. The algorithm is inspired by the behavior of real ants, which are able to find the shortest path between a food source and their nest using deposits of pheromone as a communication agent. The underlying basic self-organizing principle is exploited for the construction of parsimonious QSAR models based on neural networks for several classical QSAR data sets.

A novel method for building regression tree models for QSAR based on artificial ant colony systems.

Among the multitude of learning algorithms that can be employed for deriving quantitative structure-activity relationships, regression trees have the advantage of being able to handle large data sets, dynamically perform the key feature selection, and yield readily interpretable models. A conventional method of building a regression tree model is recursive partitioning, a fast greedy algorithm that works well in many, but not all, cases. This work introduces a novel method of data partitioning based on artificial ants. This method is shown to perform better than recursive partitioning on three well-studied data sets.

Steered molecular dynamics simulation of the Rieske subunit motion in the cytochrome bc(1) complex.

Crystallographic structures of the mitochondrial ubiquinol/cytochrome c oxidoreductase (cytochrome bc(1) complex) suggest that the mechanism of quinol oxidation by the bc(1) complex involves a substantial movement of the soluble head of the Rieske iron-sulfur protein (ISP) between reaction domains in cytochrome b and cytochrome c(1) subunits. In this paper we report the results of steered molecular dynamics simulations inducing, through an applied torque within 1 ns, a 56 degrees rotation of the soluble domain of ISP. For this purpose, a solvated structure of the bc(1) complex in a phospholipid bilayer (a total of 206,720 atoms) was constructed. A subset of 91,061 atoms was actually simulated with 45,131 moving atoms. Point charge distributions for the force field parametrization of heme groups and the Fe(2)S(2) cluster of the Rieske protein included in the simulated complex were determined. The simulations showed that rotation of the soluble domain of ISP is actually feasible. Several metastable conformations of the ISP during its rotation were identified and the interactions stabilizing the initial, final, and intermediate positions of the soluble head of the ISP domain were characterized. A pathway for proton conduction from the Q(o) site to the solvent via a water channel has been identified.

Unbinding of retinoic acid from its receptor studied by steered molecular dynamics.

Retinoic acid receptor (RAR) is a ligand-dependent transcription factor that regulates the expression of genes involved in cell growth, differentiation, and development. Binding of the retinoic acid hormone to RAR is accompanied by conformational changes in the protein which induce transactivation or transrepression of the target genes. In this paper we present a study of the hormone binding/unbinding process in order to clarify the role of some of the amino acid contacts and identify possible pathways of the all-trans retinoic acid binding/unbinding to/from human retinoic acid receptor (hRAR)-gamma. Three possible pathways were explored using steered molecular dynamics simulations. Unbinding was induced on a time scale of 1 ns by applying external forces to the hormone. The simulations suggest that the hormone may employ one pathway for binding and an alternative "back door" pathway for unbinding.

Binding pathway of retinal to bacterio-opsin: a prediction by molecular dynamics simulations.

Formation of bacteriorhodopsin (bR) from apoprotein and retinal has been studied experimentally, but the actual pathway, including the point of entry, is little understood. Molecular dynamics simulations provide a surprisingly clear prediction. A window between bR helices E and F in the transmembrane part of the protein can be identified as an entry point for retinal. Steered molecular dynamics, performed by applying a series of external forces in the range of 200-1000 pN over a period of 0.2 ns to retinal, allows one to extract this chromophore from bR once the Schiff base bond to Lys216 is cleaved. Extraction proceeds until the retinal tail forms a hydrogen bond network with Ala144, Met145, and Ser183 side groups lining the exit/entry window. The manipulation induces a distortion with a fitted root mean square deviation of coordinates (ignoring retinal, water, and hydrogen atoms) of less than 1.9 A by the time the retinal carbonyl reaches the protein surface. The forces needed to extract retinal are due to friction and do not indicate significant potential barriers. The simulations therefore suggest a pathway for the binding of retinal. Water molecules are found to play a crucial role in the binding process.

Reconstructing potential energy functions from simulated force-induced unbinding processes.

One-dimensional stochastic models demonstrate that molecular dynamics simulations of a few nanoseconds can be used to reconstruct the essential features of the binding potential of macromolecules. This can be accomplished by inducing the unbinding with the help of external forces applied to the molecules, and discounting the irreversible work performed on the system by these forces. The fluctuation-dissipation theorem sets a fundamental limit on the precision with which the binding potential can be reconstructed by this method. The uncertainty in the resulting potential is linearly proportional to the irreversible component of work performed on the system during the simulation. These results provide an a priori estimate of the energy barriers observable in molecular dynamics simulations.

Molecular dynamics study of unbinding of the avidin-biotin complex.

We report molecular dynamics simulations that induce, over periods of 40-500 ps, the unbinding of biotin from avidin by means of external harmonic forces with force constants close to those of AFM cantilevers. The applied forces are sufficiently large to reduce the overall binding energy enough to yield unbinding within the measurement time. Our study complements earlier work on biotin-streptavidin that employed a much larger harmonic force constant. The simulations reveal a variety of unbinding pathways, the role of key residues contributing to adhesion as well as the spatial range over which avidin binds biotin. In contrast to the previous studies, the calculated rupture forces exceed by far those observed. We demonstrate, in the framework of models expressed in terms of one-dimensional Langevin equations with a schematic binding potential, the associated Smoluchowski equations, and the theory of first passage times, that picosecond to nanosecond simulation of ligand unbinding requires such strong forces that the resulting protein-ligand motion proceeds far from the thermally activated regime of millisecond AFM experiments, and that simulated unbinding cannot be readily extrapolated to the experimentally observed rupture.