Prediction of Pharmaceutical Crystal Structures: a Quantum Espresso implementation with Genetic Algorithms.

FERRARO MB; LUND AM ; PAGOLA GI ; ORENDT AM; FACELLI JC

Santiagi de Chile

Congreso; WATOC 2014; 2014

UNIVERSIDAD CATOLICA de CHILE

It was shown that crystal structure prediction (CSP) based on modified genetic algorithms (MGAC [1]) coupled with universal force field methods could consistently find experimental structures of crystals. In our previous work (MGAC/CHARMM) the evaluation of the energy was made using the CHARMM [3, 4] software using the General Amber Force Field (GAFF) [5] force field. However, in many instances the inaccuracies in the force field potentials resulted in poor energetic ranking of the experimental structure or even its elimination from the candidate list limiting the usefulness of the method. To remediate this problem a new version of MGAC has been developed. The new version uses dispersion corrected density functional theory for locally optimizing the crystal structure and evaluating the energy of candidates using the software package Quantum Espresso [2] (MGAC/QE). We have determined the best choices of running parameters for crystal optimization using QE to perform local crystal structure optimizations on a test set of archetypical pharmaceutical molecules [6]. Here we present a comparison of the results obtained with the two versions of MGAC (MGAC/CHARMM and MGAC/QE) in histamine (C5H9N3). We will present the analysis of the predicted energy ranking for both methods showing that the use of QE greatly improves the energetic ranking of structures, such that the experimental match is found with a higher degree of accuracy.