Molecular dynamics (MD) simulation is a powerful “white-box” tool (defined as a tool with transparent and understandable internal mechanisms) that can reproduce, monitor and predict natural processes. Typically considered a “black box” or a “gray-box” tool (a tool with unknown or nontransparent internal mechanisms), “artificial intelligence” (AI, such as regression models) has helped with the parametrization, featurization and generalization of MD simulations in its early days. In recent years, deep learning methods have shown they have the potential to assist protein-structure predictions, nucleic acid research, small molecule designs, material designs, etc. While MD simulations excel at interpretability and data generation, they typically lack efficiency and scale. AI methods sometimes lack interpretability but could complimentarily provide efficiency and scale with the data generated by simulations and experiments. Therefore, it is of great value to build interfaces between novel deep learning technologies and MD simulations that could benefit both areas of interest.
With this Research Topic, we aim to focus on the development of methods that can bridge AI and MD methodologies. We approach the problem with two questions: (1) how do we smoothly couple AI and MD methods; and (2) how do we systematize or automate AI-MD workflows? We are asking authors to develop readily usable methods to featurize MD simulations and/or ports of these features that connect AI algorithms and MD tools. Preferably, these tools should be able to integrate with common simulation tools such as GROMACS, NAMD, LAMMPS, OpenMM, CHARMM, etc. Alternatively, authors are encouraged to design systematic or automatable workflows that help force field parametrization, spatial or temporal coarse-graining, molecular designs, featurization (such as free energy calculations), structure preparations, etc. We believe that normalizing the use of novel deep learning tools in MD simulations will assist researchers in exploiting both powerful methods in emerging fields.
We welcome submissions covering, but not limited to, the following areas:
• AI methods that featurize MD simulation data (correlation analysis, cluster analysis, rare event analysis, free energy calculations, ensemble analysis, etc.)
• Enhanced sampling using AI methods
• Structural predictions based on both AI and MD methods
• Property predictions based on both AI and MD methods
• AI-enhanced parametrization of force fields
• Automation of MD simulations with AI methods
• Toolkits that connect MD generated data and AI methods
Dr. Leili Zhang is a full time employee at IBM. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Molecular dynamics (MD) simulation is a powerful “white-box” tool (defined as a tool with transparent and understandable internal mechanisms) that can reproduce, monitor and predict natural processes. Typically considered a “black box” or a “gray-box” tool (a tool with unknown or nontransparent internal mechanisms), “artificial intelligence” (AI, such as regression models) has helped with the parametrization, featurization and generalization of MD simulations in its early days. In recent years, deep learning methods have shown they have the potential to assist protein-structure predictions, nucleic acid research, small molecule designs, material designs, etc. While MD simulations excel at interpretability and data generation, they typically lack efficiency and scale. AI methods sometimes lack interpretability but could complimentarily provide efficiency and scale with the data generated by simulations and experiments. Therefore, it is of great value to build interfaces between novel deep learning technologies and MD simulations that could benefit both areas of interest.
With this Research Topic, we aim to focus on the development of methods that can bridge AI and MD methodologies. We approach the problem with two questions: (1) how do we smoothly couple AI and MD methods; and (2) how do we systematize or automate AI-MD workflows? We are asking authors to develop readily usable methods to featurize MD simulations and/or ports of these features that connect AI algorithms and MD tools. Preferably, these tools should be able to integrate with common simulation tools such as GROMACS, NAMD, LAMMPS, OpenMM, CHARMM, etc. Alternatively, authors are encouraged to design systematic or automatable workflows that help force field parametrization, spatial or temporal coarse-graining, molecular designs, featurization (such as free energy calculations), structure preparations, etc. We believe that normalizing the use of novel deep learning tools in MD simulations will assist researchers in exploiting both powerful methods in emerging fields.
We welcome submissions covering, but not limited to, the following areas:
• AI methods that featurize MD simulation data (correlation analysis, cluster analysis, rare event analysis, free energy calculations, ensemble analysis, etc.)
• Enhanced sampling using AI methods
• Structural predictions based on both AI and MD methods
• Property predictions based on both AI and MD methods
• AI-enhanced parametrization of force fields
• Automation of MD simulations with AI methods
• Toolkits that connect MD generated data and AI methods
Dr. Leili Zhang is a full time employee at IBM. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
