Metals exert a pivotal role for the function of many biological systems. For example, metals in proteins and nucleic acids are used to assure structural stability, select appropriate conformational rearrangements, and confer specific functionality to protein-families, allowing catalytic mechanisms via the metal-aided binding of substrates. Often, metals are also crucial in therapeutic applications, being for instance components of drugs, like in the case of the metallodrug Cis-platin, i.e. the gold-standard to treat several types of cancer (Riccardi et al. NatRevChem 2018, 2). Yet, the computational investigation of metal-mediated processes poses several methodological challenges that remain only partially addressed. At the same time, metal-aided functionalities are increasingly reported in the literature, with metal ions often embedded in macromolecular assemblies composed of proteins and/or nucleic acids. It is therefore critical and timely to better understand, at atomic level, the role of metals in biological and synthetic systems, and drugs.
Indeed, the key role of metals in biology and medicine has inspired a long-lasting interest for the computational chemistry community. However, the theoretical description of metals and their realistic biological surroundings is inherently complex (Valdez et al. AccChemRes 2014, 47, 3110). The limitations of the current classical and quantum mechanical methodologies most often hamper an appropriate and complete description of mechanisms underlying metal ion-mediated biological functions.

For example, it is well known that classical force fields (FFs) are often insufficient to properly account the metal-ligand interaction in classical molecular dynamics (MD) simulations (Li & Merz, ChemRev 2017, 117). This issue is known to become severe in the case of nucleic acids and particularly for RNA, where the Mg2+ ions are critical for folding and catalysis (Casalino et al. JCTC 2017, 13, 340, Palermo et al. AccChemRes 2015, 48). Seeking for novel methodological schemes, machine-learning tools have recently been harnessed to derive classical parameters (Fracchia et al. JCTC 2018, 14, 255). As well, the emergence of polarizable force fields hold promises to improve the description of the metal-ligand interaction with the surrounding water and protein environment (Jing et al. AnnuRevBiophys 2019, 48). Catalysis, charge transfer and redox processes in biological systems require an inclusive description of the biomolecule through the use of mixed quantum-classical (QM/MM) approaches, providing a dynamic view of the biological process and the associated energetics. In this respect, the QM region is usually described via density functional theory (DFT) for its cost-effective inclusion of correlation effects. This approach is often plagued by the accuracy of the exchange correlation functionals (XCF), which calls for novel theoretical development (Brémond et al. JPhysChemLett, 2015, 6). A data-driven machine learning approach has been also put forward to overcome deficiencies of traditional XCFs in current use (Brockherde NatComm 2017, 8).

Moreover, ab-initio QM/MM approaches require massive computational resources. To overcome this limitation, a novel multi-scale framework is becoming available, which promises to exploit the power of emerging exascale computing (Olsen et al. JCTC 2019, 15), providing improved sampling and accuracy. A final methodological breakthrough is represented by the emergence of a polarizable QM/MM, which will take in consideration the effect of polarization on the QM part (Loco et al. ChemSci 2019, in press).

These methodological advances can effectively demonstrate their power only when applied to timely questions in science. In this context, we plan to discuss how the recent progress of these methods has opened new avenues for a more accurate, realistic and quantitative description of the important metal-mediated functions in biology, medicine and bioengineering.