Computational Chemistry: Reviews of Current Trends

The following sections are included:

• PREFACE

• CONTENTS

Relativistic atomic and molecular structure calculations have reached a high level of sophistication, based on algebraic approximations to the bound electronic states This review surveys basis set expansion methods for computing atomic and molecular electronic structures based on the relativistic quantum mechanics of many-body systems. We focus in particular on the important aspects of contemporary research on use of basis set methods in matrix Dirac-Fock, relativistic many-body perturbation, and coupled cluster theories.

We describe a variety of new methods for accelerating construction of the Coulomb matrix that are based upon a simplified representation of the electron charge density. Central to these methods are the elimination of degeneracies, the reduction of complexity and a Hermite Gaussian representation. The utility of Hermite Gaussian-type functions as intermediates in the computation of the Coulomb matrix is stressed, and important consequences of this representation are pointed out, especially relationships to fast numerical methods.

Projection methods are reviewed and new results obtained with two-center potential-matched fits of the electron charge density are presented. Based on these results, and on simple arguments, we conclude that the use of a single projection method to approximate the density is inefficient and introduces errors that cannot be controlled systematically.

Hierarchical multipole methods and their quantum chemical adaptations are also reviewed. We introduce the fast Gauss transform to quantum chemistry and demonstrate its equivalence to a Cartesian multipole expansion in the asymptotic limit. A synthesis of the fast Gauss transform and a Barnes-Hut-like method is suggested and a first implementation involving only the Barnes-Hut method is described and shown to be highly competitive. In particular, calculations on large polypeptides and water clusters scale as favorably as N^1,6, where N is the number of basis functions. An important feature of this new method is that accuracy is not sacrificed for speed, and that errors may be systematically controlled with a single thresholding parameter.

The introduction of the additive fuzzy electron density fragmentation principle within the ab initio Hartree-Fock quantum chemical computational framework has provided the means for generating molecular fragment densities characteristic to their local environment within a molecule. The additive fuzzy electron density fragmentation principle applied with the MEDLA (Molecular Electron Density Lego Assembler) numerical electron density fragment database, has also opened the door for the first computations of ab initio quality electron densities for macromolecules, including proteins of more than a thousand atoms. For more efficient density calculations involving small variations of the nuclear geometry, the MEDLA method has been replaced by the more accurate ALDA method (Adjustable Local Density Assembler), based on the ALDA fragment density matrix database and small readjustments of the relative positions of the atomic orbitals. The closely related ADMA (Adjustable Density Matrix Assembler) method is used to generate ab initio quality approximate density matrices for macromolecules, that allows one to apply most of the density matrix techniques of conventional quantum chemistry. A natural combination of these applications of the additive fuzzy fragmentation principle is the local shape analysis of various groups within macromolecules, providing a new method for the study of local interactions. The fundamentals of local shape analysis and similarity measures of macromolecules are discussed.

The polarizable continuum models for representing a solvent are widely used in quantum chemical computations. In this chapter we recall the current analysis of the solute solvent free energy of interaction and give the physical bases used to evalutate the electrostatic, cavitation and dispersion terms. The Self Consistent Reaction Field equations are developed, with a special emphasis on the formalism of the reaction field factors extended to the multicentric expansion of the solute's charge density and arbitrary shape cavity. The principles of post-Hartree-Fock and analytical energy derivatives computations are given.

The methods at the ab initio and semi-empirical levels are compared for some selected examples dealing with solvation free energy, modification of molecular properties induced by the solvent and reaction paths in the liquid state.

The following sections are included:

• Pentahapto Bonding

• Tying the Inverted Umbrellas

• Stability of B₁₂H₁₂²⁻ and the Structure of Elemental Boron

• Space Requirements and Stuffing of C₆₀

• Stuffing Si₆₀

• Summary

• Acknowledgements

• References

In this study, we summarize our recent ab initio calculations on the structure and interactions of DNA bases. The calculations reveal that the amino groups of isolated DNA bases adopt a nonplanar, pyramidal geometry. The MP2 calculations predict larger amino-group nonplanarity, as compared to the HF calculations. Among the DNA bases, guanine exhibits the largest nonplanarity. This intrinsic amino-group nonplanarity and flexibility contribute to the stabilization of various interactions in DNA. Base stacking and H-bonding in cytosine dimer was studied with inclusion of electron correlation at the MP2 level of theory. The H-bonded structure is more stable and is dominated by the HF interaction energy. The stability of the stacked pair originates in the electron correlation; however, the mutual orientation of the stacked bases is again due to the HF contribution. The MP2 stacking interaction energies can be reproduced quite well by the standard point-charge model with atomic charges fitted to reproduce the Molecular Electric Potential, combined with an appropriate van der Waals empirical potential. The use of additional charges below and above the molecular planes does not change the calculated interaction energies. The Distributed Multipole Analysis gives worse results than the standard point-charge model. The calculations ruled out any significant attraction originating in the interactions of polar exocyclic groups of DNA bases with the aromatic rings of bases. The stacked geometries observed in crystal structures of DNA constituents were seen to frequently correspond to very unfavorable stacking arrangements.

Traditional drug design processes have attempted to maximize drug activity with little regard for drug safely. Such manipulation usually produces highly potent derivatives with equally elevated toxicities and it results in no or disadvantageous changes in the therapeutic index. Inclusion of metabolic and toxicological considerations with the drug design process is embodied in retrometabolic concepts. This systemic methodology employs several rules for the design of safe drugs by either metabolic activation (chemical delivery system) or strategic enzymatic deactivation processes (soft drugs). The chemical delivery systems are primarily used to allow targeting of the active biological molecules to the target sites or organs. The soft drug approach is used to design new drugs by building in the molecule, in addition to the activity, the most desired way in which the molecule is to be deactivated and detoxified subsequent to exerting its biological effects. Among the various soft drug design strategies, the “inactive metabolite” and the “soft analog” approaches are the most useful and successful for designing safe and selective drugs. The rules applied to soft drug design could be extended to reduce environmental toxicity problems. Automation of these drug design processes was accomplished by computer programs. All the new generated drug candidates are conformationally optimized and ranked based on their calculated properties such as solubility, partition coefficient, isosteric-isoelectronic comparison to the lead compound, and the estimated metabolic rates of the predicted enzymatic degradation. In order to provide the best ranking information between the new drug candidates, the accurate and powerful calculated physical properties of the new drug candidates are essential. As part of this development, we have developed BLOGP and BLOGW programs with regression analysis and with the neural-net approach to calculate the partition coefficient and water solubility of a molecule based on its AM1 optimized minimum conformer.

The following sections are included:

• INDEX