De novo prediction of the ground state structure of transition metal complexes using semiempirical and ab initio quantum mechanics. Coordination isomerism

Automated Design of Realistic Organometallic Molecules from Fragments

“…169,170 The It has been suggested that a cascade of increasing accuracy molecular modeling step (molecular mechanic, semiempirical, and quantum mechanic modeling) can provide access to accurate geometries. [171][172][173][174] Nevertheless, this approach assumes that the empirical components, i.e., robust force fields and semiempirical parametrization, are well trained for the specific chemistry at hand, which is seldom the case for peculiar transition metal species, 175 thus making re-parametrization a mandatory preliminary step.…”

“…Methods for the automated generation of molecular models, either as complete structures with three-dimensional coordinates or containing constitutional information only, have mainly been developed for organic drug-like molecules in the context of in silico screening and de novo drug design . Although these methods are widely used in organic drug design, their impact in inorganic and organometallic chemistry has so far been limited. − The main reason for this difference has to do with the increased complexity of inorganic and organometallic species, i.e., the diversity of chemical bonds and flexibility with respect to both electronic and molecular structure. − In addition, the most common chemical representations are often not able to handle features particular to organometallic compounds, such as dative bonds and π-complexes …”

An Evolutionary Algorithm for de Novo Optimization of Functional Transition Metal Compounds

“…complexes using a hierarchy of molecular-level computational methods has been reported. 34 Similarly, automated molecular builders handling subclasses of metal chelate complexes also exist. 35,36 It is also encouraging that much progress has been made in the use of molecular-level computational methods for prediction of homogeneous catalysts in recent years, with a clear tendency toward the adoption of powerful quantitative structure−activity relationships (QSAR) and other computerized techniques to get the most out of the data.…”

“…Whereas high-throughput approaches and genetic algorithms have indeed been adopted in optimization of catalysts (e.g., see refs − ), the coupling of such methods with automated molecular builders to allow for incremental in silico construction and optimization of general transition metal compounds against a computed scoring (fitness) function (a hallmark of de novo design methods) has, to our knowledge, not been achieved. A de novo approach to predict the ground state coordination isomers of a subclass (trigonal bipyramidal) of transition metal complexes using a hierarchy of molecular-level computational methods has been reported . Similarly, automated molecular builders handling subclasses of metal chelate complexes also exist. , It is also encouraging that much progress has been made in the use of molecular-level computational methods for prediction of homogeneous catalysts in recent years, with a clear tendency toward the adoption of powerful quantitative structure–activity relationships (QSAR) and other computerized techniques to get the most out of the data. − All of these advances together suggest that the field is ready to explore the potential of a fully integrated drug-design-like de novo approach to design of catalysts and other molecular inorganic compounds.…”

Automated Building of Organometallic Complexes from 3D Fragments

“…A number of methods and standalone tools have been developed in the context of organic drug design for the automated generation of 3D models from representations of lower dimensionality. − Both freely available and commercial tools have been systematically tested and validated against selected data sets of organic drug-like molecules. − In contrast, the complexity of organometallic and transition-metal chemistry has hampered the application of tools for automated generation of 3D models to this class of compounds . Alternative approaches for the preparation and evaluation of 3D models have included development of specific force field methods, for instance capable of accounting for the ligand field stabilization energy, − or combination of sequences of empirical and semiempirical methods before ab initio calculations. ,− In addition, virtual construction of new 3D models has been achieved also by direct connection of structural fragments in 3D space and used in de novo design of host–guest systems and in virtual fragment-based drug design, , which is also integrated in popular software packages. , While few of the automated methods originally developed for the generation of 3D structures of organic molecules have been demonstrated to work also for specific organometallic systems, , systematic analyses of their performance with respect to organometallic and transition-metal compounds have been reported only in a couple of recent contributions. , Baldi and co-workers evaluated the performance of widely available 3D generation tools pointing out issues of accuracy and coverage for the existing methods with respect to the metal-containing molecules . To overcome these problems, they developed a new algorithm for the prediction of 3D molecular geometries, termed COSMOS, with improved performances for organometallic compounds …”