Supporting data for the paper:

"Everything everywhere all at once, a probability-based enhanced sampling approach to rare events"

Enrico Trizio, Peilin Kang and Michele Parrinello

The training of the models was based on the mlcolvar library, where the updated relevant code and updated example notebooks are available:

• didactical example: Muller-brown
• practical example: Alanine with distances

The results of the paper were obtained using the 1.2.2 version of the library.

The code for the application of the Kolmogorov bias in PLUMED is available in the file pytorch_model_bias.cpp

---

Repo structure

The contents of the repository are organized as follows:

muller : files for Muller-Brown 2D potential using PLUMED ves_md_linearexpansion

In the main text, data from iter_2_long, iter_2_long_opes_only are reported

• unbiased_sims: input files (md_input, md_potential and plumed.dat) and results (COLVAR) of the unbiased simulations used in the boundary conditions labeled dataset (A and B folders, respectively)
• biased_sims: biased simulations inputs and results from different iterations
  • generate_and_run_sims.sh: bash script to set up and run simulations for a given iteration with plumed ves_md_linearexpansion starting from a template folder (see below)
  • template: template folder for biased simulations, including the input files for starting simulations from A and B basins (input_md_A.dat, input_md_B.dat, input_md-potential.dat, plumed.dat)
  • iter_*: input files (input_md_A.dat, input_md_B.dat, input_md-potential.dat, plumed.dat) and results (A/COLVAR and B/COLVAR) obtained from simulations at different iterations of the procedure. COLVAR files have been used to incrementally build the variational loss dataset.
  • iter_2_long*: input files (input_md_A.dat, input_md_B.dat, input_md-potential.dat, plumed.dat) and results (A/COLVAR and B/COLVAR) obtained from longer production simulations at the end of the iterative procedure. COLVAR files have been used to obtain the results reported in the main text.
• models: frozen torchscript models for committor (model_*_q.pt files) and for z (model_*_z.pt files) for the different iterations

alanine : files for Alanine dipeptide in vacuum using Gromacs

In the main text, data from iter_4, iter_4_opes_only and ref_phi_psi are reported.

• unbiased_sims: unbiased simulations data used in the boundary conditions labeled dataset (COLVAR_A and COLVAR_B files)
• biased_sims: biased simulations inputs and results from different iterations
  • generate_and_run_sims.sh: bash script to set up and run simulations for a given iteration with Gromacs starting from a template folder (see below)
  • template: template folder for biased simulations, including the input files for starting simulations from A and B basins (input_A.tpr, input_B.tpr, plumed.dat)
  • iter_*: input files (input_A.tpr, input_B.tpr, plumed.dat) and results (A/COLVAR and B/COLVAR) obtained from simulations at different iterations of the procedure. COLVAR files have been used to incrementally build the variational loss dataset.
  • ref_phi_psi: files for reference OPES simulation using Phi and Psi as CVs
• models: frozen torchscript models for committor (model_*_q.pt files) and for z (model_*_z.pt files) for the different iterations

two_paths_potential : files for double path 2D potential using PLUMED ves_md_linearexpansion

In the main text, data from iter_2 are reported

• unbiased_sims: input files (input_md.dat, input_md-potential.dat and plumed.dat) and results (COLVAR) of the unbiased simulations used in the boundary conditions labeled dataset (A and B folders, respectively)
• biased_sims: biased simulations inputs and results from different iterations
  • generate_and_run_sims.sh: bash script to set up and run simulations for a given iteration with plumed ves_md_linearexpansion starting from a template folder (see below)
  • template: template folder for biased simulations, including the input files for starting simulations from A and B basins (input_md_A.dat, input_md_B.dat, input_md-potential.dat, plumed.dat)
  • iter_*: input files (input_md_A.dat, input_md_B.dat, input_md-potential.dat, plumed.dat) and results (A/COLVAR and B/COLVAR) obtained from simulations at different iterations of the procedure. COLVAR files have been used to incrementally build the variational loss dataset.
• models: frozen torchscript models for committor (model_*_q.pt files) and for z (model_*_z.pt files) for the different iterations

chignolin : files for chignolin folding in explicit water using Gromacs

Training data are too large for Github, they will be available in the Zenodo repo associated with the published paper.

• models: frozen torchscript models for committor (model_*.pt, model_*.para files) and for z (model_*_zeta.pt files) for the different iterations
• sims/template: template folder for biased and unbiased simulations, including files for starting simulations from folded and unfolded states (folded.gro, unfolded.gro, mdT.mdp, topol.top), PLUMED input files (plumed.dat, plumed-descriptors.dat, plumed_position.dat) and the additional PLUMED files (chignolin-ref.pdb for molinfo and chignolin-ca.pdb for RMSD calculation)
• models: frozen torchscript models for committor (model_iter*_q.pt, model_iter*.para files) and for z (model_iter*_z.pt files) for the different iterations
• sims/template: template folder for biased and unbiased simulations, including files for starting simulations from bound and unbound states (b.gro, u.gro, NVT.mdp, topol.top), PLUMED input files for unbiased and biased simulations (plumed.dat, plumed_biased.dat) and the additional PLUMED files (conf_template.pdb for alignment and TEST_positions.cpp for printing only atoms within the water coordination points neighborlist)

calixarene : files for OAMe-G2 binding in explicit water using Gromacs

• models: frozen torchscript models for committor (model_*.pt, model_*.para files) and for z (model_*_zeta.pt files) for the different iterations
• sims/template: template folder for biased and unbiased simulations, including files for starting simulations from bound and unbound states (b.gro, u.gro, NVT.mdp, topol.top), PLUMED input files for unbiased and biased simulations (plumed.dat, plumed_biased.dat) and the additional PLUMED files (conf_template.pdb for alignment and TEST_positions.cpp for printing only atoms within the water coordination points neighborlist)