First step:

Run infrequent metadynamics using the plumed.dat file. The basic idea of infrequent metadynamics is that the transition state along CV should be bias free. Please see https://pubs.acs.org/doi/10.1021/ct500040r for theoretical description of the method.

In our case we will use \chi_1 and \chi_2 angle of Tyrosine as CV. Our aim is to calculate transition time associated with normal to flip transition as shown in the following picture.

The flipping is defined -2.6 > \chi_1 > 2.6 which corresponds to the following code in the plumed.dat

COMMITTOR ...
  ARG=chi1_78
  STRIDE=100
  # Flipped Basin 1
  BASIN_LL1=-3.14
  BASIN_UL1=-2.6
  # Flipped basin 2
  BASIN_LL2=2.6
  BASIN_UL2=3.14
  FILE=commit.log FMT=%8.4f
... COMMITTOR

The simulation will stop once you reach the desired basin. See the COMMITTOR syntax here: https://www.plumed.org/doc-v2.6/user-doc/html/_c_o_m_m_i_t_t_o_r.html see the metadynamics syntax here:https://www.plumed.org/doc-v2.6/user-doc/html/_m_e_t_a_d.html

The idea is to launch multiple infrequent metadynamics simulations to ggather statistics. The simulation will produce a file called COLVAR-tyr whose header will look like the following

#! FIELDS time chi1_78 chi2_78 metad.bias metad.rbias metad.rct metad.work metad.acc
#! SET min_chi1_78 -pi
#! SET max_chi1_78 pi
#! SET min_chi2_78 -pi
#! SET max_chi2_78 pi
 0.000000 -1.130508 -1.245711 0.000000 -0.000000 0.000000 0.000000 1.000000

the tail will look like following

 16100.000000 -2.334819 1.604114 0.000000 -0.723659 0.723776 881.955366 2.927465
 16101.000000 -2.652049 1.922205 0.000000 -0.723776 0.723776 882.554972 2.927346

You can see when it switch to flip basin the value of metad.bias is zero

One the simulations are finished you exceute analysis.sh script which uses ts_detect.py and it prints the value of transition time in each simulation in sec

Analysis

I have gathered all the values from my simulation as transition-time.xvg and times.txt inside the ks_test and error-analysis folder respectively. transition-time.xvg and times.txt contains same values. You can rename it.

Install Matlab. Go to ks_test folder and execute the fileread.m file and you will gnerate a CDF plot and other statistics as described here: https://pubs.acs.org/doi/10.1021/ct500040r

Next we will plot the estimated p-value and transition times with error bars. Go to error-analysis folder and run error_Nsim.m. This will generate a file SD_Nsim.txt. Use error-plt.plt to plot the results

Cite the following:

1. Assessing the Reliability of the Dynamics Reconstructed from Metadynamics, Matteo Salvalaglio, Pratyush Tiwary, and Michele Parrinello, Journal of Chemical Theory and Computation 2014 10 (4), 1420-1425, DOI: 10.1021/ct500040r
2. P. Tiwary and M. Parrinello, From Metadynamics to Dynamics, Phys. Rev. Lett. 111 (230602) 2013.Link: https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.111.230602