Speaker
Description
Most of the methods used in the modelling of biomolecular systems rely on empirical parametrization and thus - at least indirectly - on some sort of training, limiting insights as well as applicability domain. A significant part of potential targets for drug design, like RNA or metalloproteins, are outside their usual applicability domain and require significant efforts of parametrization.
CCSD(T) is a generally applicable method, which is known to provide highly accurate results for all types of organic, main-group and transition metal-containing systems. The recently developed DLPNO-CCSD(T) method is an approximation to CCSD(T), providing similar accuracy at a significantly lower cost and scaling, and it is nowadays applied in routine calculations on systems with dozens to hundreds of atoms.
In this talk we present recent advances in the DLPNO-CCSD(T) method, which significantly improve its performance and accuracy, including e.g. extrapolation to the complete PNO space limit. We provide best practices for its application in real-life (bio)chemical applications. We show recent applications of the DLPNO-CCSD(T) method in biomolecular systems, including protein-ligand interactions as well RNA-ligand interactions. The computed accurate interaction energy can be decomposed into its components between different molecular fragments, and we show how this can be applied in deciphering key interactions between a ligand and its biomolecular target.
