Jakub Lengiewicz

Video recording:

Abstract: Machine Learning Force Fields (MLFFs) enable the modeling of chemical systems by combining ab initio accuracy with the efficiency of classical force fields. Despite the great success of ML methods, extending their applicability to larger molecules poses a challenge, partly due to the rapid growth of the dimensionality of the descriptor. State-of-the-art descriptors include non-essential degrees of freedom or neglect long-range interactions by imposing a cut-off radius. Thus, finding a way to accurately describe both short- and long-range interactions without significantly increasing descriptor size is a critical step required to advance ML modeling.

In this seminar, I will discuss recent progress and challenges for next-generation MLFFs. Specifically, I will focus on global MLFFs that can efficiently model large and flexible molecules without resorting to any potentially uncontrolled approximation [1]. I will demonstrate the possibility of achieving linear scaling in global MLFFs for large systems through an automated descriptor reduction approach [2]. The studied systems include units of four major types of biomolecules and a supramolecular complex.

[1] doi.org/10.1126/sciadv.adf0873

[2] doi.org/10.1038/s41467-023-39214-w

Adil Kabylda is a PhD Researcher in Theoretical Chemical Physics Group at the University of Luxembourg. He obtained B.Sc. and M.Sc. degrees in Chemistry from Moscow State University in 2021. His research focuses on extending the applicability of Machine Learning Force Fields to larger (bio)molecules, with a particular emphasis on accurately describing long-range interactions.