covariance matrix (only with a correlation matrix in the case of
a normalized PCA).
- In addition to the 2D principal component plots, other types
of graphic plot are of interest and provide for example an
insight on protein regions whose movements contribute the
most to explain the conformational diversity along each princi-
pal component (residue wise loadings plot). All these analyses
are well described in the section PCA of the online tutorial
using bio3d R package at http://thegrantlab.org/bio3d/
tutorials/trajectory-analysis. - It should be taken into account that the energy used by
reweight script is in kT units. In Amber 14 the energy unit is
kcal/mol and must be converted. - One has to be sure that the considered system has reached its
equilibrium state before introducing a bias in the simulation.
Indeed, if the system is not equilibrated enough, the estimated
free energy variation could be the combination of the none-
quilibrated part of the system and the bias introduced. - The greater the force constant k, the smaller the region
explored by the system. This results in narrower distributions
ofζand therefore a smaller overlap between two successive
windows. On the contrary, the lower the force constantk,the
greater the region explored by the system. This can have the
effect of exploring several stable states and therefore obtaining
non-Gaussian multimodal distributions ofζ. - The smaller the step size between windows, the greater the
sampling. However the number of windows strongly impacts
the amount of computing resources needed.
Acknowledgments
This work was supported by the Institut de Recherche Servier and
the French National Research Agency (ANR-13-JSV5-0001 and
ANR-15-CE20-0015). The authors wish to thank the Re ́gion
Centre Val de Loire and the Ligue contre le Cancer for financial
supports and the Orle ́ans-Tours CaSciModOT at the Centre de
Calcul Scientique de la Re ́gion Centre Val de Loire and the Centre
Re ́gional Informatique et d’Applications Nume ́riques de Norman-
die (CRIANN) for providing computer facilities.
References
- Du X, Li Y, Xia Y-L et al (2016) Insights into
protein–ligand interactions: mechanisms,
models, and methods. Int J Mol Sci 17:144.
https://doi.org/10.3390/ijms17020144
2. Changeux JP, Edelstein S (2011) Conforma-
tional selection or induced fit? 50 years of
debate resolved. F1000 Biol Rep.https://doi.
org/10.3410/B3-19
424 Sonia Ziada et al.