An analysis of the structural and energetic properties of deoxyribose by potential energy methods

We discuss the three fundamental issues of a computational approach in structure prediction by potential energy minimization, and analyze them for the nucleic acid component deoxyribose. Predicting the conformation of deoxyribose is important not only because of the molecule's central conformational role in the nucleotide backbone, but also because energetic and geometric discrepancies from experimental data have exposed some underlying uncertainties in potential energy calculations. The three fundamental issues examined here are: (i) choice of coordinate system to represent the molecular conformation; (ii) construction of the potential energy function; and (iii) choice of the minimization technique. For our study, we use the following combination. First, the molecular conformation is represented in cartesian coordinate space with the full set of degrees of freedom. This provides an opportunity for comparison with the pseudorotation approximation. Second, the potential energy function is constructed so that all the interactions other than the nonbonded terms are represented by polynomials of the coordinate variables. Third, two powerful Newton methods that are globally and quadratically convergent are implemented: Gill and Murray's Modified Newton method and a Truncated Newton method, specifically developed for potential energy minimization. These strategies have produced the two experimentally-observed structures of deoxyribose with geometric data (bond angles and dihedral angles) in very good agreement with experiment. More generally, the application of these modeling and minimization techniques to potential energy investigations is promising. The use of cartesian variables and polynomial representation of bond length, bond angle and torsional potentials promotes efficient second-derivative computation and, hence, application of Newton methods. The truncated Newton, in particular, is ideally suited for potential energy minimization not only because the storage and computational requirements of Newton methods are made manageable, but also because it contains an important algorithmic adaptive feature: the minimization search is diverted from regions where the function is nonconvex and is directed quickly toward physically interesting regions.     

Krylov subspace accelerated inexact Newton method for linear and nonlinear equations

A Krylov subspace accelerated inexact Newton (KAIN) method for solving linear and nonlinear equations is described, and its relationship to the popular direct inversion in the iterative subspace method [DIIS; Pulay, P., Chem Phys Lett 1980, 393, 73] is analyzed. The two methods are compared with application to simple test equations and the location of the minimum energy crossing point of potential energy surfaces. KAIN is no more complicated to implement than DIIS, but can accommodate a wider variety of preconditioning and performs substantially better with poor preconditioning. With perfect preconditioning, KAIN is shown to be very similar to DIIS. For these reasons, KAIN is recommended as a replacement for DIIS.     

Second- and higher-order convergence in linear and nonlinear multiconfigurational Hartree–Fock theory

We discuss how the local convergence of Newton–Raphson and fixed Hessian MCSCF iterative models may be rationalized in terms of a total order of convergence in an error vector and a corresponding error term. We demonstrate that a sequence of N Newton–Raphson iterations has a total order of convergence of 2^N and that a sequence of N fixed Hessian iterations has a total order of convergence of N + 1. We derive the error terms of a Newton–Raphson and a fixed Hessian sequence of iterations. We discuss the implementation of the fixed Hessian and the Newton–Raphson approaches both when linear and nonlinear transformations of the variables are carried out. Sample calculations show that insight into the structure of the local convergence of Newton–Raphson and fixed Hessian models can be based on an order of convergence and an error term analysis.     

Calculation of wave‐functions with frozen orbitals in mixed quantum mechanics/molecular mechanics methods. II. Application of the local basis equation

The application of the local basis equation (Ferenczy and Adams, J. Chem. Phys. 2009 , 130, 134108) in mixed quantum mechanics/molecular mechanics (QM/MM) and quantum mechanics/quantum mechanics (QM/QM) methods is investigated. This equation is suitable to derive local basis nonorthogonal orbitals that minimize the energy of the system and it exhibits good convergence properties in a self‐consistent field solution. These features make the equation appropriate to be used in mixed QM/MM and QM/QM methods to optimize orbitals in the field of frozen localized orbitals connecting the subsystems. Calculations performed for several properties in divers systems show that the method is robust with various choices of the frozen orbitals and frontier atom properties. With appropriate basis set assignment, it gives results equivalent with those of a related approach [G. G. Ferenczy previous paper in this issue] using the Huzinaga equation. Thus, the local basis equation can be used in mixed QM/MM methods with small size quantum subsystems to calculate properties in good agreement with reference Hartree–Fock–Roothaan results. It is shown that bond charges are not necessary when the local basis equation is applied, although they are required for the self‐consistent field solution of the Huzinaga equation based method. Conversely, the deformation of the wave‐function near to the boundary is observed without bond charges and this has a significant effect on deprotonation energies but a less pronounced effect when the total charge of the system is conserved. The local basis equation can also be used to define a two layer quantum system with nonorthogonal localized orbitals surrounding the central delocalized quantum subsystem. © 2013 Wiley Periodicals, Inc.     

A truncated Newton minimizer adapted for CHARMM and biomolecular applications

We report the adaptation of the truncated Newton minimization package TNPACK for CHARMM and biomolecular energy minimization. TNPACK is based on the preconditioned linear conjugate–gradient technique for solving the Newton equations. The structure of the problem—sparsity of the Hessian—is exploited for preconditioning. Experience with the new version of TNPACK is presented on a series of molecular systems of biological and numerical interest: alanine dipeptide (N-methyl-alanyl-acetamide), a dimer of N-methyl-acetamide, deca-alanine, mellitin (26 residues), avian pancreatic polypeptide (36 residues), rubredoxin (52 residues), bovine pancreatic trypsin inhibitor (58 residues), a dimer of insulin (99 residues), and lysozyme (130 residues). Detailed comparisons among the minimization algorithms available in CHARMM, particularly those used for large-scale problems, are presented along with new mathematical developments in TNPACK. The new TNPACK version performs significantly better than ABNR, the most competitive minimizer in CHARMM, for all systems tested in terms of CPU time when curvature information (Hessian/vector product) is calculated by a finite-difference of gradients (the numeric option of TNPACK). The remaining derivative quantities are, however, evaluated analytically in TNPACK. The CPU gain is 50% or more (speedup factors of 1.5 to 2.5) for the largest molecular systems tested and even greater for smaller systems (CPU factors of 1 to 4 for small systems and 1 to 5 for medium systems). TNPACK uses curvature information to escape from undesired configurational regions and to ensure the identification of true local minima. It converges rapidly once a convex region is reached and achieves very low final gradient norms, such as of order 10^?8, with little additional work. Even greater overall CPU gains are expected for large-scale minimization problems by making the architectures of CHARMM and TNPACK more compatible with respect to the second-derivative calculations. © 1994 by John Wiley & Sons, Inc.     

Ball convergence of a sixth-order Newton-like method based on means under weak conditions

We study the local convergence of a Newton-like method of convergence order six to approximate a locally unique solution of a nonlinear equation. Earlier studies show convergence under hypotheses on the seventh derivative or even higher. The convergence in this study is shown under hypotheses on the first derivative although only the first derivative appears in these methods. Hence, the applicability of the method is expanded. Finally, we solve the problem of the fractional conversion in the ammonia process showing the applicability of the theoretical results.     

Optimization of quantum Monte Carlo wave functions by energy minimization

We study three wave function optimization methods based on energy minimization in a variational Monte Carlo framework: the Newton, linear, and perturbative methods. In the Newton method, the parameter variations are calculated from the energy gradient and Hessian, using a reduced variance statistical estimator for the latter. In the linear method, the parameter variations are found by diagonalizing a nonsymmetric estimator of the Hamiltonian matrix in the space spanned by the wave function and its derivatives with respect to the parameters, making use of a strong zero-variance principle. In the less computationally expensive perturbative method, the parameter variations are calculated by approximately solving the generalized eigenvalue equation of the linear method by a nonorthogonal perturbation theory. These general methods are illustrated here by the optimization of wave functions consisting of a Jastrow factor multiplied by an expansion in configuration state functions (CSFs) for the C2 molecule, including both valence and core electrons in the calculation. The Newton and linear methods are very efficient for the optimization of the Jastrow, CSF, and orbital parameters. The perturbative method is a good alternative for the optimization of just the CSF and orbital parameters. Although the optimization is performed at the variational Monte Carlo level, we observe for the C2 molecule studied here, and for other systems we have studied, that as more parameters in the trial wave functions are optimized, the diffusion Monte Carlo total energy improves monotonically, implying that the nodal hypersurface also improves monotonically.     

Convergence of Newton’s method under Vertgeim conditions: new extensions using restricted convergence domains

We present new sufficient convergence conditions for the semilocal convergence of Newton’s method to a locally unique solution of a nonlinear equation in a Banach space. We use Hölder and center Hölder conditions, instead of just Hölder conditions, for the first derivative of the operator involved in combination with our new idea of restricted convergence domains. This way, we find a more precise location where the iterates lie, leading to at least as small Hölder constants as in earlier studies. The new convergence conditions are weaker, the error bounds are tighter and the information on the solution at least as precise as before. These advantages are obtained under the same computational cost. Numerical examples show that our results can be used to solve equations where older results cannot.     

Linear-scaling implementation of molecular electronic self-consistent field theory

A linear-scaling implementation of Hartree-Fock and Kohn-Sham self-consistent field (SCF) theories is presented and illustrated with applications to molecules consisting of more than 1000 atoms. The diagonalization bottleneck of traditional SCF methods is avoided by carrying out a minimization of the Roothaan-Hall (RH) energy function and solving the Newton equations using the preconditioned conjugate-gradient (PCG) method. For rapid PCG convergence, the Lowdin orthogonal atomic orbital basis is used. The resulting linear-scaling trust-region Roothaan-Hall (LS-TRRH) method works by the introduction of a level-shift parameter in the RH Newton equations. A great advantage of the LS-TRRH method is that the optimal level shift can be determined at no extra cost, ensuring fast and robust convergence of both the SCF iterations and the level-shifted Newton equations. For density averaging, the authors use the trust-region density-subspace minimization (TRDSM) method, which, unlike the traditional direct inversion in the iterative subspace (DIIS) scheme, is firmly based on the principle of energy minimization. When combined with a linear-scaling evaluation of the Fock/Kohn-Sham matrix (including a boxed fitting of the electron density), LS-TRRH and TRDSM methods constitute the linear-scaling trust-region SCF (LS-TRSCF) method. The LS-TRSCF method compares favorably with the traditional SCF/DIIS scheme, converging smoothly and reliably in cases where the latter method fails. In one case where the LS-TRSCF method converges smoothly to a minimum, the SCF/DIIS method converges to a saddle point.     

Performance of hybrid methods for large-scale unconstrained optimization as applied to models of proteins

Energy minimization plays an important role in structure determination and analysis of proteins, peptides, and other organic molecules; therefore, development of efficient minimization algorithms is important. Recently, Morales and Nocedal developed hybrid methods for large-scale unconstrained optimization that interlace iterations of the limited-memory BFGS method (L-BFGS) and the Hessian-free Newton method (Computat Opt Appl 2002, 21, 143-154). We test the performance of this approach as compared to those of the L-BFGS algorithm of Liu and Nocedal and the truncated Newton (TN) with automatic preconditioner of Nash, as applied to the protein bovine pancreatic trypsin inhibitor (BPTI) and a loop of the protein ribonuclease A. These systems are described by the all-atom AMBER force field with a dielectric constant epsilon = 1 and a distance-dependent dielectric function epsilon = 2r, where r is the distance between two atoms. It is shown that for the optimal parameters the hybrid approach is typically two times more efficient in terms of CPU time and function/gradient calculations than the two other methods. The advantage of the hybrid approach increases as the electrostatic interactions become stronger, that is, in going from epsilon = 2r to epsilon = 1, which leads to a more rugged and probably more nonlinear potential energy surface. However, no general rule that defines the optimal parameters has been found and their determination requires a relatively large number of trial-and-error calculations for each problem.     

Conformational search using a molecular dynamics–minimization procedure: Applications to clusters of coulombic charges,Lennard–Jones particles,and waters

A new conformational search method, molecular dynamics–minimization (MDM), is proposed, which combines a molecular dynamics sampling strategy with energy minimizations in the search for low-energy molecular structures. This new method is applied to the search for low energy configurations of clusters of coulombic charges on a unit sphere, Lennard–Jones clusters, and water clusters. The MDM method is shown to be efficient in finding the lowest energy configurations of these clusters. A closer comparison of MDM with alternative conformational search methods on Lennard–Jones clusters shows that, although MDM is not as efficient as the Monte Carlo–minimization method in locating the global energy minima, it is more efficient than the diffusion equation method and the method of minimization from randomly generated structures. Given the versatility of the molecular dynamics sampling strategy in comparison to Monte Carlo in treating molecular complexes or molecules in explicit solution, one anticipates that the MDM method could be profitably applied to conformational search problems where the number of degrees of freedom is much greater. © 1998 John Wiley & Sons, Inc. J Comput Chem 19: 60–70, 1998     

Adaptive multilevel finite element solution of the Poisson–Boltzmann equation I. Algorithms and examples

This article is the first of two articles on the adaptive multilevel finite element treatment of the nonlinear Poisson–Boltzmann equation (PBE), a nonlinear eliptic equation arising in biomolecular modeling. Fast and accurate numerical solution of the PBE is usually difficult to accomplish, due to the presence of discontinuous coefficients, delta functions, three spatial dimensions, unbounded domain, and rapid (exponential) nonlinearity. In this first article, we explain how adaptive multilevel finite element methods can be used to obtain extremely accurate solutions to the PBE with very modest computational resources, and we present some illustrative examples using two well‐known test problems. The PBE is first discretized with piece‐wise linear finite elements over a very coarse simplex triangulation of the domain. The resulting nonlinear algebraic equations are solved with global inexact Newton methods, which we have described in an article appearing previously in this journal. A posteriori error estimates are then computed from this discrete solution, which then drives a simplex subdivision algorithm for performing adaptive mesh refinement. The discretize–solve–estimate–refine procedure is then repeated, until a nearly uniform solution quality is obtained. The sequence of unstructured meshes is used to apply multilevel methods in conjunction with global inexact Newton methods, so that the cost of solving the nonlinear algebraic equations at each step approaches optimal O(N) linear complexity. All of the numerical procedures are implemented in MANIFOLD CODE (MC), a computer program designed and built by the first author over several years at Caltech and UC San Diego. MC is designed to solve a very general class of nonlinear elliptic equations on complicated domains in two and three dimensions. We describe some of the key features of MC, and give a detailed analysis of its performance for two model PBE problems, with comparisons to the alternative methods. It is shown that the best available uniform mesh‐based finite difference or box‐method algorithms, including multilevel methods, require substantially more time to reach a target PBE solution accuracy than the adaptive multilevel methods in MC. In the second article, we develop an error estimator based on geometric solvent accessibility, and present a series of detailed numerical experiments for several complex biomolecules. © 2000 John Wiley & Sons, Inc. J Comput Chem 21: 1319–1342, 2000     

Secant-like methods for solving nonlinear models with applications to chemistry

We present a local as well a semilocal convergence analysis of secant-like methods under g eneral conditions in order to approximate a locally unique solution of a nonlinear equation in a Banach space setting. The new conditions are more flexible than in earlier studies. This way we expand the applicability of these methods, since the new convergence conditions are weaker. Moreover, these advantages are obtained under the same conditions as in earlier studies. Numerical examples are also provided in this study, where our results compare favorably to earlier ones.     

Solutions to linear matrix ordinary differential equations via minimal,regular, and excessive space extension based universalization

This work focuses on the solution of the linear matrix ordinary differential equations where the first derivative of the unknown matrix is equal to the same unknown matrix premultiplied by a given matrix polynomially varying with the independent variable. Work aims to get a universal form for this equation by using the space extension concept where new unknowns are defined to get more amenable form for the equation. The convergence of the series solution to this equation obtained via minimal, regular, and excessive space extension is also investigated with the aid of an appropriate norm analysis which also enables us to get error estimates for the truncated series solutions. A few illustrative examples are presented for practical convergence issues like approximation quality.     

An efficient algorithm for solving nonlinear equations with a minimal number of trial vectors: applications to atomic-orbital based coupled-cluster theory

The conjugate residual with optimal trial vectors (CROP) algorithm is developed. In this algorithm, the optimal trial vectors of the iterations are used as basis vectors in the iterative subspace. For linear equations and nonlinear equations with a small-to-medium nonlinearity, the iterative subspace may be truncated to a three-dimensional subspace with no or little loss of convergence rate, and the norm of the residual decreases in each iteration. The efficiency of the algorithm is demonstrated by solving the equations of coupled-cluster theory with single and double excitations in the atomic orbital basis. By performing calculations on H(2)O with various bond lengths, the algorithm is tested for varying degrees of nonlinearity. In general, the CROP algorithm with a three-dimensional subspace exhibits fast and stable convergence and outperforms the standard direct inversion in iterative subspace method.     

Analysis of the genetic algorithm method of molecular conformation determination

Many important problems in chemistry require knowledge of the 3-D conformation of a molecule. A commonly used computational approach is to search for a variety of low-energy conformations. Here, we study the behavior of the genetic algorithm (GA) method as a global search technique for finding these low-energy conformations. Our test molecule is cyclic hexaglycine. The goal of this study is to determine how to best utilize GAs to find low-energy populations of conformations given a fixed amount of CPU time. Two measures are presented that help monitor the improvement in the GA populations and their loss of diversity. Different hybrid methods that combine coarse GA global search with local gradient minimization are evaluated. We present several specific recommendations about trade-offs when choosing GA parameters such as population size, number of generations, rate of interaction between subpopulations, and combinations of GA and gradient minimization. In particular, our results illustrate why approaches that emphasize convergence of the GA can actually decrease its effectiveness as a global conformation search method. © John Wiley & Sons, Inc.     

A new method for fast and accurate derivation of molecular conformations

During molecular simulations, three-dimensional conformations of biomolecules are calculated from the values of their bond angles, bond lengths, and torsional angles. In this paper we study how to efficiently derive three-dimensional molecular conformations from the values of torsional angles. This case is of broad interest as torsional angles greatly affect molecular shape and are always taken into account during simulations. We first review two widely used methods for deriving molecular conformations, the simple rotations scheme and the Denavit-Hartenberg local frames method. We discuss their disadvantages which include extensive bookkeeping, accumulation of numerical errors, and redundancies in the local frames used. Then we introduce a new, fast, and accurate method called the atomgroup local frames method. This new method not only eliminates the disadvantages of earlier approaches but also provides lazy evaluation of atom positions and reduces the computational cost. Our method is especially useful in applications where many conformations are generated or updated such as in energy minimization and conformational search.     

Multiparametric curve fitting XIV. Modus operandi of the least-squares algorithm MINOPT

Hybrid least-squares algorithm MINOPT for a nonlinear regression is introduced. MINOPT from CHEMSTAT package combines fast convergence of the Gauss-Newton method in a vicinity of minimum with good convergence of gradient methods for location far from a minimum. Quality of minimization and an accuracy of parameter estimates for six selected models are examined and compared with different derivative least-squares methods of five commercial regression packages.     

Remarks on the analysis of torsional energy surfaces of cycloheptane and cyclooctane by molecular mechanics

A modified version (MM 2′) of the Allinger's 1977 force field is checked against cycloheptane and cyclooctane. Cycloheptane is characterized by two pseudorotating itineraries, chair/twist-chair and boat/twist-boat, separated by a barrier of 8.5 kcal mol^?1. The activation energy in the C/TC pseudorotation is estimated to be 0.96 kcal mol^?1, while B and TB transform into each other freely at an energy level 3.8 kcal mol^?1 above the global energy minimum (TC). With cyclooctane the lowest energy is calculated for the boat-chair form which participates in a pseudorotational process with TBC through a saddle point lying 3.5 kcal mol^?1 above BC. The chair/chair and boat/boat families contain only one local minimum, crown and BB, respectively, on the MM 2′ surface. The results are presented as an illustration for quick coverage of torsional energy surface by two-bond driver calculation with the block-diagonal Newton–Raphson minimization, followed by the force search of stationary points by full-matrix Newton–Raphson optimization.