On the second day of the summer school, Prof. Cecilia Clementi (Rice University) gave a lecture on learning effective molecular models using machine learning methods. In particular, she addressed the problem to capture rare events – like protein folding – in molecular dynamics simulations and how to achieve long-time simulations using “short-time” data.
In the second lecture, Connor Coley (MIT) gave an introduction to using machine learning for synthetic chemistry. He demonstrated that reaction pathways for producing medical drugs can be optimized or even discovered from data available from the literature or databases.
Prof. Noé (FU Berlin) provided an overview on (artificial) neural networks and addressed some potential pitfalls. In the second part of his lecture, he returned to the topic capturing rare events in molecular dynamics simulations. Markov models and novel machine learning approaches allow it to realize very long-time simulations.
In the hand-on session, Prof. French (Case Western University) introduced the programming environment/language “R” which is popular among data scientists.
The day was concluded with a poster session where the participants showed and discussed their research results.