An alternative to the “Star Path” enhancement of the ADMA linear scaling method for protein modeling

Abstract

With the aim of improving the performance of macromolecular quantum chemistry conformation analysis and reaction path following methods, the Adjustable Density Matrix Assembler (ADMA) method has already been combined with some faster although less accurate density matrix extrapolation methods, such as the Löwdin-Inverse-Löwdin (LIL) extrapolation along a potential energy surface, and a strategically arranged back-and-forth switching between these methods has been proven to be advantageous. Here, an alternative approach is proposed and investigated, based on several actual test calculations, where the "inexpensive" LIL density matrix extrapolation steps are replaced by only somewhat more expensive, but still ADMA-based calculations, where in the "rough-search stage," only interactions of shorter distances within the macromolecule are considered. It is shown that this approach is viable, as an alternative to the "Star Path" method including both ADMA and LIL steps. © 2017 Wiley Periodicals, Inc.

Analysis of two overlapping fragmentation approaches in density matrix construction: GMBE-DM vs. ADMA

“…In all GMBE-DM and ADMA calculations below, each water molecule or each ion component with an integer charge is considered as a fragment. ADMA has been employed exclusively with small and nondiffuse basis sets, including STO-3G, 8–10,15 3-21G, 10,11 6-31G, 11 6-31G(d), 13 6-31G(d,p), 7–9 6-311G, 10,12 and 6-311G(d). 10 For large and higher-quality diffuse basis sets, a very large distance cutoff is essential to generate sufficiently large parent molecules for ADMA.…”

“…Considering the challenges mentioned, various alternative methods based on chemical fragmentation aim to provide high quality initial guesses for supersystem calculations by utilizing subsystem density matrices. These include the adjustable density matrix assembler (ADMA) approach, [6][7][8][9][10][11][12][13][14][15] the molecular orbitals of the fragment molecular orbital (FMO-MO) method, [16][17][18][19] the molecular fractionation with conjugate caps-assembled density matrix (MFCC-DM) method, 5,20,21 the divide and conquer (DC) approach, 22,23 and our approach based on the many-body expansion for building density matrices (MBE-DM). 1 All these approaches follow a two-step fragmentation process: first, assembling the supersystem density matrix, and then computing a property related to that density matrix, typically the total energy.…”