Parameter optimization on the convergence surface of PATH simulations Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
  • Chandrasekaran, Srinivas Niranj
    • Affiliation: School of Medicine, Department of Biochemistry and Biophysics
  • Computational treatments of protein conformational changes tend to focus on the trajectories themselves, despite the fact that it is the transition state structures that contain information about the barriers that impose multi-state behavior. PATH is an algorithm that computes a transition pathway between two protein crystal structures, along with the transition state structure, by minimizing the Onsager-Machlup action functional. It is rapid but depends on several unknown input parameters whose range of different values can potentially generate different transition-state structures. Transition-state structures arising from different input parameters cannot be uniquely compared with those generated by other methods. I outline modifications that I have made to the PATH algorithm that estimates these input parameters in a manner that circumvents these difficulties, and describe two complementary tests that validate the transition-state structures found by the PATH algorithm. First, I show that although the PATH algorithm and two other approaches to computing transition pathways produce different low-energy structures connecting the initial and final ground-states with the transition state, all three methods agree closely on the configurations of their transition states. Second, I show that the PATH transition states are close to the saddle points of free-energy surfaces connecting initial and final states generated by replica-exchange Discrete Molecular Dynamics simulations. I show that aromatic side-chain rearrangements create similar potential energy barriers in the transition-state structures identified by PATH for a signaling protein, a contractile protein, and an enzyme. Finally, I observed, but cannot account for, the fact that trajectories obtained for all-atom and Cα-only simulations identify transition state structures in which the Cα atoms are in essentially the same positions. The consistency between transition-state structures derived by different algorithms for unrelated protein systems argues that although functionally important protein conformational change trajectories are to a degree stochastic, they nonetheless pass through a well-defined transition state whose detailed structural properties can rapidly be identified using PATH. In the end, I outline the strategies that could enhance the efficiency and applicability of PATH.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Hermans, Jan
  • Dokholyan, Nikolay
  • Zhang, Qi
  • Carter, Charles
  • Campbell, Sharon
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2016

This work has no parents.