Accelerated molecular dynamics (aMD)

Accelerated molecular dynamics (aMD) simulations present another possible approach

to overcoming the issue of systems being trapped in local potential free energy minima. This
method aims to increase the frequency of transitions across potential energy barriers by
introducing a bias potential such that the potential surfaces near minima are raised, while those
near the barriers or saddle points are left unmodified, thus decreasing the barrier heightsintact
(as shown in Figure aMD1). [Hamelberg 2004] No prior knowledge about the system (such as
the need to pre-select a CV in metadynamics) is required for aMD.Compared to umbrella
sampling, [Torrie 1977] another well-known enhanced sampling technique based on the addition
of a compensating function to the true potential, aMD, much like metadynamics, [Laio 2008]
does not require prior knowledge of conformations of interest. In contrast, aMD introduces an
energy boost at potential energy wells, defined as areas where the system’s potential energy
falls below a chosen threshold. The added bias potential, ∆V, depends on the system’s potential
energy V: is defined as

{
( )❑❑ ❑
ΔV ( r )= ❑❑❑ ❑
()❑❑ ❑
❑❑
where E Vthreshold is the threshold boost energy, V(r) is the original potential, and 𝛼 is a tuning
parameter that determines the depth of the potential energymodified potential well. This
definition expression offor ∆V ensures that the bias potential mimics the shape of the underlying
potential minima., and ensuring convergence to a Boltzmann distribution of conformations (Tthe
exact original potential free energy landscape can then be reconstructed using Boltzmann
reweighting., [Miao 2014] One noteworthy caveat is that aMD is not a fully standalone method –
the boosting parameters α and Vthreshold are system-specific and have to be tuned, which is
typically done based on the number of atoms or residues, as well as average energies from
prior conventional MD runs. [Pierce 2012] One must also be careful when choosing the
reweighting algorithm, as exponential average reweighting is known to introduce high energetic
fluctuations, due to a small number of high-boost potential frames dominating the reweighting
factors. [Miao 2014]This problem can be avoided by using Maclaurin expansion [Pierce 2012],
though this expansion does not always give the correct energy minimum positions. [Miao 2014]
Cumulant expansion [Shen 2008, Park 2003, Jarzynski 1997] tends to avoid these issues and
tends to be the best choice for a wide array of biomolecules. [Miao 2014] [uh here you need to
discuss the caveats of aMD. For example, need to run cMD to see what you should choose for
alpha and Vthreshold (? Is this true? What the heck is alpha?). also introduce the “energetic
noise” problem that can happen during reweighting (?)]with a computational load that compares
favorably to conventional MD. While aMD is marginally more computationally expensive per
step than conventional MD, on average it tends to represent more effective simulation time.
[Hamelberg 2004]
 Figure aMD1. A schematic representation of accelerated MD, showing the true potential
(solid line), the biased potential (dashed line), and the threshold boost energy (dotted line).
Adopted from Hamelberg 2004.

To test whether aMD simulations with an explicit solvent model could sample the
conformational landscape of constrained cyclic peptides, Kamenik et al. utilized aMD to explore
the conformational landscape ofused three small cyclic peptides with potential therapeutic
useas benchmarks (Figure aMD2): cyclo-(PSlDVPro-Ser-leu-Asp-Val), cyclo-(RGDfV), and
cyclo-(RRWWRF) (Figure aMD2; in the peptide sequences, small lowercase letters denote D-
amino acids and underline denotes N-methylatedion). All three cyclic peptides have potential
therapeutic applications: Cyclo-(PSlDV) is an integrin binder,;[refsViles 1996] cyclo-(RGDfV),
also called cilengitide, is a potential anticancer drug;,[refsXiong 2002, Moruno 2010, Marelli
2014] and cyclo-(RRWWRF) is an antimicrobial peptide.[refsAppelt 2005] Moreover, solution
NMR studies had been reported for all three cyclic peptides.[Viles 1996, Marelli 2014, Appelt
2005] It was found that in cyclo-(PSlDV),NMR analysis of cyclo-(PSlDV) found up to five
conformers in slow exchange. The two most abundant components were found to have a cis-
and trans- amide bond the amide bond between Pro1 and Val5, respectively, with all other
bonds being trans. adopted a cis:trans distribution of ??:?? in water, and there were up to five
conformers in slow exchangeTheir relative abundance was found to be 16% trans- to 66% cis.
[Viles 1996] No cis peptide bonds were observed in the NMR studies of the other two peptides.
[Marelli 2014, Appelt 2005] cyclo-(RRWWRF) was determined to be highly flexible in aqueous
solution, only assuming a more rigid structure when bound to a micelle.[Appelt 2005][describe
key NMR findings for cilengitide and cyclo-(RRWWRF) that u’ll compare the simulation results
with later here]
an integrin binder which is known to occupy states consistent with both cis- and trans-
Pro, [Viles 1996] cilengitide (cyclo-(Arg-Gly-Asp-phe-[NMe-Val])), a previously mentioned cyclic
pentapeptide that has been extensively studied as a potential anticancer drug, [Marelli 2014]
and cyclo-(Arg-Arg-Trp-Trp-Arg-Phe), an antimicrobial CP drug. [Appelt 2005] [Kamenik 2018]
The small size of these molecules makes simulations less computationally expensive. This, in
combination with being experimentally well-characterized and diverse in terms of flexibility,
makes them a good test bed for evaluating the efficiency and accuracy of aMD as an enhanced
sampling method.

The authors conducted theAccelerated MD simulations were carried out using the
AMBERff14SB force field [Maier 2015] with the TIP3P water model. [Jorgensen 1983] In
the case of cilengitide, additional parameters for N-methylated amino acids were taken
from Forcefield_NCAA. [Khoury 2014] The simulations were performed at atmospheric
pressure and a temperature of 300 K. One-microsecondOne1- μs trajectories of both
conventional MD and both conventional MD and aMD were collected for cyclo-(PSlDV) and
cyclo-(RRWWRF); for cilengitide, the authors performed twenty 50 ns simulations, combining
them into a 1 μs trajectory. All aMD simulations used the dual-boost algorithm implemented in
the Amber16 [Case 2016] software package – in other words, a bias was applied to the total
potential, along with an additional bias component on the dihedral term. [Hamelberg 2004,
Hamelberg 2007][uh you need to mention this dual boost scheme used in their aMD…] and The
gGlobal structural ensemble was analyzed using dihedral PCA and cluster analysis,
following a Boltzmann reweighting process [Miao 2014] to recover the unbiased
ensembleresults. For cilengitide, the authors performed twenty 50 ns simulations, combining
them into a 1 μs trajectory. In order to make data analysis faster and more robust, they
cluster analysis was performed focused on the 2,000 snapshots (0.4% of all trajectory
snapshots) with the highest boosting potential, i.e. the lowest potential energy. [Kamenik
2018]

Kamenik et al. first showed that while a 1-μs It was determined that conventional MD
simulations exhibited limited sampling of cyclo-(PSlDV), aMD was able to explore both the cis
and trans isomers for the amide bond between Pro1 and Val5 (15)were not able to capture
significant structural changes in the first model system over a 1 μs trajectory, demonstrating the
efficiency of aMD sampling. [Kamenik 2018] The cis:trans distribution of 15 in the reweighted
aMD results was 25:75, highly consistent with thatwhich is consistent with the ratio reported in
the solution NMR studiesy.[Viles 1996] To gauge accuracy, aMD was benchmarked against
experimentally collected NMR data. For cyclo-PSlDV, the 2000 snapshots with the highest
boosting potential yielded a cis/trans distribution of 25:75, which was in good agreement with
the experimentally determined distribution of 20:80. [Viles 1996] Using the entire ensemble, for
comparison, again yielded the same 25:75 distribution. All average interproton distance
restraints for the trans structures were satisfied;, whereas for the cis structures, 17 out of
18 NOE restraints were met with one distance found to be 0.2 Å too small in the simulation.
The cluster analysis showed cyclo-(PSlDV) adopted that three distinct conformationsstates
(one trans and two cis) were densely sampled by aMD, which was a good match for NMR-
derived dataconsistent with NMR findings. For cilengitide, all NOE restraints but one were
met, with a small violation of 0.24 Å. The one violation observed was suggested by authors to
be due to a force field inaccuracy related to the presence of D-Phe in the molecule. [Kamenik
2018] aMD was able to sample a minimum corresponding to the bioactive conformation of
cilengitide, with a representative structure having a backbone RMSD of 0.6 Å compared to the
known bound conformation. The conformational landscape of cyclo-(RRWWRF)cyclo-
RRWWRF exhibited a large number of minima, which is, again, consistent with the NMR
results tha, whicht suggested XXXthat this cyclic peptide was highly flexible in aqueous
solution. [Appelt 2005] All NOE restraints were met for average interproton distances;
however, individual structures still show large violations, further supporting that the
experimental NOE distances resulted from averaging. While aMD simulations seemed to
reproduced the experimental NOE results, Lastly, the 3J𝛼NH3JαNH coupling constants for cyclo-
(PSlDV) determined from aMD were not found to be in good agreement with NMR-derived
ones., This discrepancy suggestsing that aMD does not provide a high resolution when it
comes to subtle differences between dihedral distributions. This [add a noun here ah ah ah
ah ah~~~!!!] is, likely due to a large amount of “energetic noise” introduced in the reweighting
steppotentially as a result of energetic noise introduced in the reweighting step. [Kamenik
2018]

Figure aMD2. [new figure draft is available as a Google Slide in this folder] The cyclic
peptides studied by Kamenik et al. (1) cyclo-PSlDV, an integrin binder in which the
marked peptide bond distinguishes the cis- and trans- states, (2)
Cilengitide, an anticancer drug candidate. (3) cyclo-RRWWRF, an
antimicrobial hexapeptide. Adopted from Kamenik 2018.