Biological systems are inherently multiscale. Molecular interactions and biochemical reactivity are governed by electronic rearrangements, whereas collective processes, such as protein folding, span much larger spatial and temporal scales. Modeling such processes requires computational approaches that combine accuracy and efficiency across multiple resolutions. Multiscale modeling, and in particular quantum mechanics/molecular mechanics (QM/MM) methods, provide a powerful technique to investigate such phenomena. However, their high computational cost limits their applicability to small regions treated at higher resolution and for relatively short time scales.
This thesis focuses on the application and development of multiscale modeling methods, with particular emphasis on applications to biological systems containing transition metal complexes. In the first part, QM/MM methods were employed to support and rationalize experimental findings in an investigation of ruthenium- and osmium-based anticancer compounds targeting histone proteins in the nucleosome core particle.
When the binding site is unknown, QM/MM approaches are computationally prohibitive for large-scale screening of many different candidate locations. To address this, we developed Metal3D and Metal1D, two predictive tools for identifying metal binding sites in proteins. Initially trained on zinc ions, they also demonstrated predictive ability for other transition metal ions, and we evaluated their applicability to identify binding sites for transition metal based complexes coordinating proteins in a mono- or bi-dentate fashion. While not yet fully generally applicable, both tools represent a first step toward efficient binding site identification, which can then be refined through QM/MM simulations.
We then focused on contributing to the development of the MiMiC framework for multiscale simulations. MiMiC couples external programs assigned to different subsystems (e.g., one subsystem treated at the QM level and another with MM), and it computes the resulting interactions. In this work, the framework was extended by interfacing new external programs, enabling novel combinations for QM/MM molecular dynamics simulations.
Furthermore, new features were introduced within MiMiC itself, including a scheme for dynamically fitting atom-centered electric multipole moments during QM/MM molecular dynamics simulations.
Overall, this thesis demonstrates both the power and limitations of multiscale modeling approaches. The contributions presented here provide a foundation for more reliable and versatile computational strategies to study transition metal based agents in complex biomolecular contexts, with potential applications in drug design, catalysis, and beyond.
