Force fields
My expertise in force fields spans the development of machine learning force fields and the parameterization of classical force fields, enabling accurate and efficient molecular simulations.
I have contributed significantly to developing machine learning force fields that reproduce high-quality quantum chemical calculations. My work focuses on creating robust, scalable, and highly accurate ML models that capture complex chemical interactions, dynamics, and properties. These advanced ML force fields have demonstrated remarkable efficiency and precision, particularly suited for simulations where classical methods fall short or where high quantum mechanical accuracy is crucial.
Additionally, my experience includes extensive parameterization of classical force fields, particularly for ligands, cofactors, and non-standard residues. I ensure these parameters accurately represent molecules’ chemical and physical behavior, enabling realistic biomolecular simulations and drug design studies. My parameterization procedures involve rigorous validation against experimental and high-level quantum chemical reference data, ensuring reliability and predictive power.
Relevant projects#
Many-body gradient-domain machine learning#
Atomistic insight is fundamental for computational predictive studies of chemical and physical processes. Machine learning force fields provide a route to high-level ab initio calculations at a fraction of the cost. Gradient-domain machine learning (GDML), a kernel-based method, directly learns the relationship between atomic coordinates and interatomic forces. However, training in the gradient domain sacrifices generalized transferability to other species or number of atoms.
Many-body GDML (mbGDML), is a technique for GDML transferability to n-sized systems by using many-body machine learning models. Every aspect of the process from preparing quantum chemistry calculations, data set creation, training, and use of mbGDML force fields is taken care of in this user-friendly Python package.