From macromolecules to electrons—grand challenges in theoretical and computational chemistry

Among the many achievements of the twentieth century the accelerating development of microprocessors and their capabilities is one of the most impressive accomplishments, influencing virtually every aspect of daily life. The impact of this new technological resource was (and still is) of particular importance for theoretical approaches in science and engineering. Although many theories (Schrödinger, 1926a,b; Dirac, 1928; Feynman et al., 2010) required for an accurate treatment of quantum systems such as light and matter have been formulated prior to the construction of vacuum tube computers and the invention of the transistor, the applicability of these methodologies is strongly linked to the capacities of the employed computational equipment. Initially being merely considered as a supplement to experimental work, the steady and consistent development of theoretical approaches significantly extended their accuracy and capabilities. At present many of these disciplines comprise wellestablished, independent research fields. The computational treatment of chemical systems (Allen and Tildesley, 1990; Szabo and Ostlund, 1996; Levine, 1999; Sadus, 1999; Helgaker et al., 2000; Frenkel and Smit, 2002; Cook, 2005; Dyall, 2007; Reiher and Wolf, 2009; Tuckerman, 2010) is a prominent example of this development. Although modern computational and theoretical chemistry is the result of decades of brilliant research and an impressive wealth of knowledge has been gathered, a number of topics show still as much activity as during the early phases of this discipline. A breakthrough in any of these topics would mark an exceptional milestone and therefore, these questions are considered as grand challenges in the field of theoretical and computational chemistry in the next decades. Perhaps one of the most prominent challenges in chemistry became known as the protein folding problem (Dill and MacCallum, 2012), i.e., the prediction of the folded structures of peptide and protein systems. While it is tempting to focus on a “sequence-to-structure” relationship, it was argued that such a direct translation does not exist (Ben-Naim, 2013) simply due to the fact that proteins may fold differently when exposed to different chemical environments. Although the protein folding problem may be considered as a purely biochemical topic, it is a highly interdisciplinary field of research, linking biology and biochemistry to fields such as analytical, inorganic, medicinal, organic, physical, pharmaceutical and theoretical chemistry. The latter enables the possibility to investigate the system on a microscopic (i.e., atomistic) level, whereas the vast majority of experimental approaches work in the macroscopic regime. The most critical aspect determining the accuracy of a computational treatment of chemical systems lies in the description of the atomic interactions. Despite the tremendous capabilities of modern high performance computing facilities, it is still necessary to formulate a compromise between accuracy of results and computational effort. At present pairwise-additive, non-polarizable, empirical potential approaches referred to as molecular mechanical or force field methods (Leach, 2001; Cramer, 2002; Jensen, 2006; Ramachandran et al., 2008) represent the state-of-the-art, but the question whether these approaches are sufficiently accurate has been asked on a regular basis (Hummer et al., 2008; Allison et al., 2011; Beauchamp et al., 2012). Among the many interatomic forces occurring within such systems, the accurate representation of the solute-solvent interaction is of particular difficulty. While in the past the surrounding solvent was largely regarded as necessary evil when conducting simulation studies (in some cases it was even ignored), an increased awareness of the importance of the solvation is observed in literature in recent years (Levy and Onuchic, 2004; Zaccai, 2004; De Simone et al., 2005; Mamontov and Chu, 2012; Xu et al., 2012). On the other hand the interaction between the biomacromolecule and solvent is of utmost importance, since the resulting intermolecular forces guide the protein into its folded state (Ben-Naim, 2013). Research focused on an accurate description of biomolecular solvation can be regarded as one grand challenge, which due to the sheer size of proteins and the associated number of molecules involved in the solvation process is by no means a simple task. Since the respective potential models are of empirical nature and experimental data delivers only macroscopic information of the investigated systems, the question arises which data serve as accurate reference on the microscopic level. Naturally, the answer to this lies in the domain of quantum chemistry (Szabo and Ostlund, 1996; Levine, 1999; Helgaker et al., 2000; Cook, 2005; Dyall, 2007; Reiher and Wolf, 2009; Cramer, 2002). In contrast to empirical models, which rely on parameterized interactions, quantum chemical methods aim to achieve an accurate description of chemical systems