BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning from molecular dynamics trajectory ensembles - Professor 
 Peter Bolhuis\, University of Amsterdam
DTSTART:20231011T133000Z
DTEND:20231011T143000Z
UID:TALK201610@talks.cam.ac.uk
CONTACT:Lisa Masters
DESCRIPTION:Molecular dynamics (MD) is a powerful computational tool with 
 applications ranging from chemical reactions\, to phase transitions\, to b
 iomolecular conformational changes. However\, in practice\, MD is far from
  fulfilling this promise due to exceedingly long simulation times caused b
 y high free energy barriers\, also known as the rare event problem.  One c
 an overcome this problem using enhanced sampling. This requires knowledge 
 of the reaction coordinate (RC): the principal collective variable or feat
 ure that determines the progress along an activated or reactive process. A
  good RC is crucial for generating sufficient statistics with enhanced sam
 pling. Moreover\, the RC provides invaluable atomistic insight in the proc
 ess under study. The optimal RC is the committor\, which can be computed w
 ith brute force MD\, or more efficiently by e.g. Transition Path Sampling 
 (TPS).  Novel schemes for TPS using reinforcement learning can now effecti
 vely map the committor function. The interpretability of the committor\, b
 eing a high dimensional function\, remains very low.  Applying dimensional
 ity reduction can reveal the RC in terms of low-dimensional human understa
 ndable molecular collective variables (CVs) or order parameters.  Here\, I
  discuss several methods to perform this dimensionality reduction [1].\nIn
  the second part\, I focus on a general framework of imposing known rate c
 onstants as constraints in molecular dynamics simulations\, based on a com
 bination of the maximum-entropy (MaxEnt) and maximum-caliber principles (M
 axCal). Starting from an existing ensemble of (rare event) dynamical traje
 ctories or paths\, e.g. obtained from TPS\, each path is reweighted in ord
 er to match the calculated and experimental interconversion rates of a mol
 ecular transition of interest\, while minimally perturbing the prior path 
 distribution [2]. This kinetically corrected ensemble of trajectories lead
 s to improved structure\, kinetics and thermodynamics.  One also learns me
 chanistic insight that may not be readily evident directly from the experi
 ments.  This method does not alter the Hamiltonian directly\, and therefor
 e we recently proposed a novel MaxCal-based path-reweighting technique to 
 optimize parameters in the molecular model itself\, while constraining kin
 etic observables [3]. This opens up the possibility to design molecular mo
 dels that lead to desired kinetic behaviour. \n\n[1] M. Frassek\, A. Arjun
 \, and P. G. Bolhuis\, J. Chem. Phys. 155\, 064103 (2021).\n[2] Z. F. Brot
 zakis\, M. Vendruscolo\, and P. G. Bolhuis\, Proc. Natl. Acad. Sci. 118\, 
 (2021).\n[3] P. G. Bolhuis\, Z. F. Brotzakis\, and B. G. Keller\, J. Chem.
  Phys. 159\, 074102 (2023) .\n
LOCATION:Unilever Lecture Theatre\, Yusuf Hamied Department of Chemistry
END:VEVENT
END:VCALENDAR