Computational Drug Discovery and Design

(backadmin) #1

  1. MARTINI is a widely used CG model but cannot be used for
    modeling backbone conformational transitions owing to the
    use of an elastic network model (ENM) to restrain protein
    secondary structure to its starting conformation [169].
    A recent development of MARTINI has been able to fold
    some protein fragments by replacing the ENM restraints with
    a contact map [170].

  2. To perform NMA on an atomistic protein model a powerful
    minimization is required (to find the bottom of the approxi-
    mated harmonic basin) followed by diagonalization of the
    second derivative of the potential energy (3 N3 N Hessian
    matrix).

  3. Recent application code libraries such as Bio3D and ProDy
    offer advanced combinations of ENMs like pathway sampling
    strategies or correlation analyses, and can be used to build
    combinations for custom pipelines [164, 165].

  4. The Gaussian height is reduced over time such that the simula-
    tion is forced to converge as the height tends to zero.

  5. There is continual exploration of the optimal arrangements for
    replica-exchange but generally exchanges are made every
    2–10,000 time steps and optimized such that at least 30% of
    exchanges are accepted [171].

  6. Temperature replica-exchange with even moderately sized,
    explicitly solvated, systems requires many replicas owing to
    the high heat capacity of water.

  7. The experiences of the current authors suggest that biasing
    over multiple replicas, each with an uncorrelated driven vari-
    able, is often able to explore protein conformational space in a
    very powerful way but can be difficult to converge.

  8. There are currently at least five academic codes for managing
    large adaptive celling calculations [172–177].

  9. A single Anton 2 computer can produce a ~115 us simulation
    per day on the standard industry test system. It should be
    noted that many important biological processes take place
    over much longer timescales. Anton 2 has been optimized for
    simulations of large models (~700,000 atoms), which fits with
    their aims of “investigating the mechanisms of several pharma-
    ceutically relevant cellular receptors, transport proteins, and
    enzymes” as some of the suggested systems of study would
    be very large.

  10. For example Morphic Therapeutics focused entirely on a very
    detailed understanding of the molecular mode of action of the
    integrin family of proteins to achieve a first in class therapeutic.


Computational Study of Protein Conformational Transitions 357
Free download pdf