"Machine learning in chemistry and beyond" (ChE-650) seminar by Sereina Riniker (ETH Zurich)
Event details
| Date | 23.11.2021 |
| Hour | 15:15 › 16:15 |
| Speaker | Sereina Riniker is currently Associate Professor of Computational Chemistry at the Department of Chemistry and Applied Biosciences of ETH Zurich. |
| Location | Online |
| Category | Conferences - Seminars |
| Event Language | English |
Using machine learning for molecular dynamics simulations
From simple clustering techniques to sophisticated neural networks, the use of machine learning has become a valuable tool in many fields of chemistry in the past decades. Here, we describe different ways in which we explore the combination of machine learning (ML) and molecular dynamics (MD) simulations. One topic focuses on how the information in MD simulations can be encoded as input to train ML models for the quantitative understanding of molecular systems. Molecular dynamics fingerprints (MDFP) represent an orthogonal description of molecules compared to topological fingerprints. The concept of the MDFPs is highly versatile, depending on the property to be predicted, different systems can be simulated and different properties can be extracted from the MD simulations. The second topic addresses the utilization of ML to improve the set-up, interpretation, as well as accuracy of MD simulations. In classical MD simulations, the physical interactions between atoms are described with an empirical force field. This involves a large number of parameters for each molecule, which are fitted to quantum-mechanical (QM) or available experimental data. There is a need for more accurate and general force fields. In this context, we demonstrate how ML approaches can aid in force-field development, e.g. for the improved generation of partial charges for organic molecules. In the third part, we explore the use of ML for increasing the speed and accuracy of QM/MM MD simulations.
From simple clustering techniques to sophisticated neural networks, the use of machine learning has become a valuable tool in many fields of chemistry in the past decades. Here, we describe different ways in which we explore the combination of machine learning (ML) and molecular dynamics (MD) simulations. One topic focuses on how the information in MD simulations can be encoded as input to train ML models for the quantitative understanding of molecular systems. Molecular dynamics fingerprints (MDFP) represent an orthogonal description of molecules compared to topological fingerprints. The concept of the MDFPs is highly versatile, depending on the property to be predicted, different systems can be simulated and different properties can be extracted from the MD simulations. The second topic addresses the utilization of ML to improve the set-up, interpretation, as well as accuracy of MD simulations. In classical MD simulations, the physical interactions between atoms are described with an empirical force field. This involves a large number of parameters for each molecule, which are fitted to quantum-mechanical (QM) or available experimental data. There is a need for more accurate and general force fields. In this context, we demonstrate how ML approaches can aid in force-field development, e.g. for the improved generation of partial charges for organic molecules. In the third part, we explore the use of ML for increasing the speed and accuracy of QM/MM MD simulations.
Practical information
- Informed public
- Free
Contact
- Kevin Maik Jablonka, Solène Oberli, Puck van Gerwen