Skolnick Research Group Publications
http://hdl.handle.net/1853/26287
Publications of Dr. Skolnick's Research Group
20190618T08:59:43Z

Krylov subspace methods for computing hydrodynamic interactions in Brownian dynamics simulations
http://hdl.handle.net/1853/45067
Krylov subspace methods for computing hydrodynamic interactions in Brownian dynamics simulations
Ando, Tadashi; Chow, Edmond; Saad, Yousef; Skolnick, Jeffrey
Hydrodynamic interactions play an important role in the dynamics of macromolecules. The most common way to take into account hydrodynamic effects in molecular simulations is in the context of a Brownian dynamics simulation. However, the calculation of correlated Brownian noise vectors in these simulations is computationally very demanding and alternative methods are desirable. This paper studies methods based on Krylov subspaces for computing Brownian noise vectors. These methods are related to Chebyshev polynomial approximations, but do not require eigenvalue estimates. We show that only low accuracy is required in the Brownian noise vectors to accurately compute values of dynamic and static properties of polymer and monodisperse suspension models. With this level of accuracy, the computational time of Krylov subspace methods scales very nearly as O(N²) for the number of particles N up to 10 000, which was the limit tested. The performance of the Krylov subspace methods, especially the “block” version, is slightly better than that of the Chebyshev method, even without taking into account the additional cost of eigenvalue estimates required by the latter. Furthermore, at N = 10 000, the Krylov subspace method is 13 times faster than the exact Cholesky method. Thus, Krylov subspace methods are recommended for performing largescale Brownian dynamics simulations with hydrodynamic interactions.
© 2012 American Institute of Physics; The electronic version of this article is the complete one and can be found online at: http://dx.doi.org/10.1063/1.4742347; DOI: 10.1063/1.4742347
20120801T00:00:00Z
Ando, Tadashi
Chow, Edmond
Saad, Yousef
Skolnick, Jeffrey
Hydrodynamic interactions play an important role in the dynamics of macromolecules. The most common way to take into account hydrodynamic effects in molecular simulations is in the context of a Brownian dynamics simulation. However, the calculation of correlated Brownian noise vectors in these simulations is computationally very demanding and alternative methods are desirable. This paper studies methods based on Krylov subspaces for computing Brownian noise vectors. These methods are related to Chebyshev polynomial approximations, but do not require eigenvalue estimates. We show that only low accuracy is required in the Brownian noise vectors to accurately compute values of dynamic and static properties of polymer and monodisperse suspension models. With this level of accuracy, the computational time of Krylov subspace methods scales very nearly as O(N²) for the number of particles N up to 10 000, which was the limit tested. The performance of the Krylov subspace methods, especially the “block” version, is slightly better than that of the Chebyshev method, even without taking into account the additional cost of eigenvalue estimates required by the latter. Furthermore, at N = 10 000, the Krylov subspace method is 13 times faster than the exact Cholesky method. Thus, Krylov subspace methods are recommended for performing largescale Brownian dynamics simulations with hydrodynamic interactions.

GOAP: A Generalized OrientationDependent, AllAtom Statistical Potential for Protein Structure Prediction
http://hdl.handle.net/1853/43439
GOAP: A Generalized OrientationDependent, AllAtom Statistical Potential for Protein Structure Prediction
Zhou, Hongyi; Skolnick, Jeffrey
An accurate scoring function is a key component for successful protein structure prediction. To address this important unsolved problem, we develop a generalized orientation and distancedependent allatom statistical potential. The new statistical potential, generalized orientationdependent allatom potential (GOAP), depends on the relative orientation of the planes associated with each heavy atom in interacting pairs. GOAP is a generalization of previous orientationdependent potentials that consider only representative atoms or blocks of sidechain or polar atoms. GOAP is decomposed into distance and angledependent contributions. The DFIRE distancescaled finite ideal gas reference state is employed for the distancedependent component of GOAP. GOAP was tested on 11 commonly used decoy sets containing 278 targets, and recognized 226 native structures as best from the decoys, whereas DFIRE recognized 127 targets. The major improvement comes from decoy sets that have homologymodeled structures that are close to native (all within ∼4.0 Å) or from the ROSETTA ab initio decoy set. For these two kinds of decoys, orientationindependent DFIRE or only sidechain orientationdependent RWplus performed poorly. Although the OPUSPSP blockbased orientationdependent, sidechain atom contact potential performs much better (recognizing 196 targets) than DFIRE, RWplus, and dDFIRE, it is still ∼15% worse than GOAP. Thus, GOAP is a promising advance in knowledgebased, allatom statistical potentials. GOAP is available for download at http://cssb.biology.gatech.edu/GOAP.
© 2011 Biophysical Society.; DOI: 10.1016/j.bpj.2011.09.012
20111001T00:00:00Z
Zhou, Hongyi
Skolnick, Jeffrey
An accurate scoring function is a key component for successful protein structure prediction. To address this important unsolved problem, we develop a generalized orientation and distancedependent allatom statistical potential. The new statistical potential, generalized orientationdependent allatom potential (GOAP), depends on the relative orientation of the planes associated with each heavy atom in interacting pairs. GOAP is a generalization of previous orientationdependent potentials that consider only representative atoms or blocks of sidechain or polar atoms. GOAP is decomposed into distance and angledependent contributions. The DFIRE distancescaled finite ideal gas reference state is employed for the distancedependent component of GOAP. GOAP was tested on 11 commonly used decoy sets containing 278 targets, and recognized 226 native structures as best from the decoys, whereas DFIRE recognized 127 targets. The major improvement comes from decoy sets that have homologymodeled structures that are close to native (all within ∼4.0 Å) or from the ROSETTA ab initio decoy set. For these two kinds of decoys, orientationindependent DFIRE or only sidechain orientationdependent RWplus performed poorly. Although the OPUSPSP blockbased orientationdependent, sidechain atom contact potential performs much better (recognizing 196 targets) than DFIRE, RWplus, and dDFIRE, it is still ∼15% worse than GOAP. Thus, GOAP is a promising advance in knowledgebased, allatom statistical potentials. GOAP is available for download at http://cssb.biology.gatech.edu/GOAP.

An object oriented environment for artificial evolution of protein sequences: The example of rational design of transmembrane sequences
http://hdl.handle.net/1853/41970
An object oriented environment for artificial evolution of protein sequences: The example of rational design of transmembrane sequences
Milik, Mariusz; Skolnick, Jeffrey
A system is presented for generating peptide sequences with desirable properties,
using combination of neural network and artificial evolution. The
process is illustrated by an example of a practical problem of generating
artificial transbilayer peptides. The peptides generated in the process of
artificial evolution have the physicochemical properties of transmembrane
peptides, and forms stable transmembrane structures in testing Monte Carlo
simulations. The artificial evolution system is designed to emulate natural
evolution; therefore it is of both practical and theoretical interest, both in
terms of rational design of protein sequences and modeling of natural evolution
of proteins.
©1995 Massachusetts Institute of Technology Press; Presented at Evolutionary Programming IV: Proceedings of the Fourth Annual Conference on Evolutionary Programming, March 13, 1995, San Diego, California.
19950101T00:00:00Z
Milik, Mariusz
Skolnick, Jeffrey
A system is presented for generating peptide sequences with desirable properties,
using combination of neural network and artificial evolution. The
process is illustrated by an example of a practical problem of generating
artificial transbilayer peptides. The peptides generated in the process of
artificial evolution have the physicochemical properties of transmembrane
peptides, and forms stable transmembrane structures in testing Monte Carlo
simulations. The artificial evolution system is designed to emulate natural
evolution; therefore it is of both practical and theoretical interest, both in
terms of rational design of protein sequences and modeling of natural evolution
of proteins.

Optimization of protein structure on lattices using a selfconsistent field approach
http://hdl.handle.net/1853/41969
Optimization of protein structure on lattices using a selfconsistent field approach
Reva, Boris A.; Rykunov, D. S.; Finkelstein, Alexei V.; Skolnick, Jeffrey
Lattice modeling of proteins is commonly used to study the protein folding problem. The reduced number of possible conformations of lattice models enormously facilitates exploration of the conformational space. In this work, we suggest a method to search for the optimal lattice models that reproduced the offlattice structures with minimal errors in geometry and energetics. The method is based on the selfconsistent field optimization of a combined pseudoenergy function that includes two force fields: an "interaction field," that drives the residues to optimize the chain energy, and a "geometrical field," that attracts the residues towards their native positions. By varying the contributions ofthese force fields in the combined pseudoenergy, one can also test the accuracy of potentials: the better the potentials, i.e., the more accurate the "interaction field," and the smaller the contribution of the "geometrical field" required for building accurate lattice models
This is a copy of an article published in the Journal of Computational Biology ©1998 Mary Ann Liebert, Inc.; Journal of Computational Biology is available online at: http://www.liebertonline.com.
19980101T00:00:00Z
Reva, Boris A.
Rykunov, D. S.
Finkelstein, Alexei V.
Skolnick, Jeffrey
Lattice modeling of proteins is commonly used to study the protein folding problem. The reduced number of possible conformations of lattice models enormously facilitates exploration of the conformational space. In this work, we suggest a method to search for the optimal lattice models that reproduced the offlattice structures with minimal errors in geometry and energetics. The method is based on the selfconsistent field optimization of a combined pseudoenergy function that includes two force fields: an "interaction field," that drives the residues to optimize the chain energy, and a "geometrical field," that attracts the residues towards their native positions. By varying the contributions ofthese force fields in the combined pseudoenergy, one can also test the accuracy of potentials: the better the potentials, i.e., the more accurate the "interaction field," and the smaller the contribution of the "geometrical field" required for building accurate lattice models