Date/Time
Date(s) - 03/03/2022
1:30 pm - 2:30 pm

We are excited to welcome Dr. Guido von Rudorff, Postdoctoral Researcher at the University of Vienna, on March 3.

We will be offering the seminar in a hybrid format (details below).

Hope to see you there!

Presenter: Dr. Guido von Rudorff, Postdoctoral Researcher, University of Vienna

Title: Design in Compound Space with Machine Learning and Quantum Alchemy

Date: Thursday, March 3, 2022

Time: 1:30-2:30

Room: ABB-162

Zoom: email chemgrad@mcmaster.ca for the link

Abstract: When searching for compounds of desirable properties, the two main challenges are the vastness of the search space and the expense to measure or calculate the properties for just one compound. Machine learning allows to speed up property evaluation by often four to five orders of magnitude by learning trends in chemical space, while quantum alchemy allows to reduce the search space by treating changes in elements perturbatively. Machine learning applications shown include predicting reaction barriers without explicit knowledge of the transition state geometry, absorption spectra for molecular identification, and obtaining 3D geometries for small organic molecules, carbenes, and transition state structures. With quantum alchemy, covalent and non-covalent interactions will be estimated four orders of magnitude faster than with simple enumeration. Additionally, a new symmetry is presented: alchemical enantiomers. Even though they may differ in bonds, alchemical enantiomers need to be quasi-degenerate. This symmetry allows to build more efficient machine-learning models and is a prime example how quantum alchemy allows us to transform design approaches to go away from one-by-one enumeration of a never-ending list of candidate compounds.