Life is physics and chemistry in action. While molecular simulations of systems as complex as whole cells are now within reach, predicting chemical reactivity on relevant time and length scales remains a challenge. I will present our recent work towards bringing action – here: chemistry – to classical simulations and molecular design through machine learning. Among others, we substitute costly quantum mechanical calculations with a graph neural network-based emulator. Our framework can deal with the plethora of life’s chemistry amidst the ‘jiggling and wiggling’ of biomolecules. Importantly, we also uncover unexpected biomolecular processes that we in turn put to test in experiments. Finally, I will demonstrate how we harness a flow-matching model to predict biomolecular dynamics. Our method paves the way for generating novel flexible and functional proteins.

Top-quark pair production in association with a jet is a key process at the LHC. Its high sensitivity to the top mass and the increasing experimental precision call for the QCD corrections to be computed at the next-to-next-to-leading order (NNLO). In this seminar, I will present the computation of the two-loop Feynman integrals required to obtain NNLO QCD predictions in the leading colour approximation. These integrals are characterised by significant algebraic complexity—stemming from the multi-scale, five-particle kinematics—as well as analytic complexity, due to the appearance of nested square roots and elliptic functions. I will discuss modern methods for tackling multi-scale integrals in a way that is suitable for phenomenology, and outline first steps to extend these techniques to cases involving elliptic functions.