reorga

README.md

@@ -18,7 +18,7 @@ The parameters of the CG lipid models are iteratively and simultaneously optimiz
 
 ![Overview of the principles used in SwarmCG for model and force field building](https://github.com/GMPavanLab/SwarmCGM/blob/main/images/Fig1.png)
 
-This code was recently applied as part of the efforts deployed for the further enhancement of lipid models in Martini 3. Proof-of-concept and testing of different CG resolutions for building alternative Martini lipid models is soon to be published in JCIM and a [preprint version is available on ChemRxiv](chemrxiv.org/engage/chemrxiv/article-details/642e9c97736114c96301544d). Resulting models obtained for phosphatidylcholine (PC) lipids are [available here](github.com/GMPavanLab/SwarmCGM/tree/main/resulting_models/). 
+This code was recently applied as part of efforts for the further enhancement of lipid models in Martini 3. Proof-of-concept and testing of different CG resolutions for building alternative Martini lipid models is soon to be published in JCIM and a [preprint version is available on ChemRxiv](chemrxiv.org/engage/chemrxiv/article-details/642e9c97736114c96301544d). Resulting models obtained for phosphatidylcholine (PC) lipids are [available here](https://github.com/GMPavanLab/SwarmCGM/tree/main/resulting_models/). 
 
 # System Requirements
 
@@ -54,15 +54,15 @@ Supported HPC resource managers:
 
 # Installation Guide
 
-Standalone files can be directly copied as [provided here](github.com/GMPavanLab/SwarmCGM/tree/main/).
+Standalone files can be directly copied as [provided here](https://github.com/GMPavanLab/SwarmCGM/tree/main/).
 
 Users must install the following python packages prior to executing the code:
 
 ```
 pip/pip3 install numpy scipy matplotlib MDAnalysis pyemd fst-pso pyyaml
 ```
 
-If you are having troubles installing or using this software, please [drop us an Issue](github.com/GMPavanLab/SwarmCGM/issues). 
+If you are having troubles installing or using this software, please [drop us an Issue](https://github.com/GMPavanLab/SwarmCGM/issues). 
 
 Tested python & package versions:
 - python: 3.6.12, 3.8.5
@@ -97,7 +97,7 @@ These values can be even further decreased for the purpose of testing the consis
 As the input of SwarmCG requires a preliminary CG mapping choice (*i.e.* defining the positions and types of the CG beads, bonds and angles used for building the molecular models which parameters will be optimized), several parameters allow to define precisely which parameters should be optimized, or which other ones stay at given fixed values, according to your requirements.
 To this end, SwarmCG makes use of [YAML](yaml.org) config files to simplify this process and help keeping track of the hyper-parameters used for an optimization run. 
 
-Please refer to [the demo YAML config file](github.com/GMPavanLab/SwarmCGM/tree/main/minimalistic_demo.yaml) for a breakdown of each parameter and step-by-step guidance on how to parametrize an optimization procedure.
+Please refer to [the demo YAML config file](https://github.com/GMPavanLab/SwarmCGM/tree/main/minimalistic_demo.yaml) for a breakdown of each parameter and step-by-step guidance on how to parametrize an optimization procedure.
 These parameters can also be related to the content and explanations provided in the paper.
 
 # Citation

minimalistic_demo.yaml

@@ -10,7 +10,7 @@ lipids_codes:  # lipids and temperatures to use for this optimization (other inp
   POPC: [303K]
   DOPC: [303K]
 map_center: COG  # COG or COM, for using Center Of Geometry or Center Of Mass when reading the NDX files and getting the AA-mapped bottom-up ref. data
-reset: True  # if True, recalculates bond and angle distributions and distance distributions between pair of beads from the AA data
+reset: True  # if True, recalculates bond and angle distributions and distance distributions between pair of beads from the AA data, otherwise, they are saved as a pickle and will be re-used
 phosphate_pos:  # position of the phosphate atom or bead, used for calculating Dhh membrane thickness
   PC:  # per head type
     AA: 20  # atom id of the phosphate, 1-indexed
@@ -72,32 +72,32 @@ gpu_ids: "X X"  # string, space-separated GPU ids to use per simulation slot def
 gmx_gpu_cancel_str: "-nb cpu"  # string, will be added to GROMACS commands for the slots which have disabled GPU usage (= marked 'X' in 'gpu_ids')
 
 # 9. INITIAL FORCE FIELD PARAMETERS FOR BONDED TERMS
-init_bonded:  # must correspond to bond and angle types as defined in your input ITP files
-  B1: { val: 0, fct: 1250 }  # Bonds
-  B2: { val: 0, fct: 1250 }  # nm, kJ.mol-1.nm-2
-  B3: { val: 0, fct: 1250 }
-  B4: { val: 0, fct: 1250 }
-  B5: { val: 0, fct: 1250 }
-  B6: { val: 0, fct: 1250 }
-  B7: { val: 0, fct: 1250 }
-  B8: { val: 0, fct: 1250 }
-  B10: { val: 0, fct: 1250 }
-  B11: { val: 0, fct: 1250 }
-  A1: { val: 0, fct: 50 }  # Angles
-  A2: { val: 0, fct: 50 }  # degrees, kJ.mol-1
-  A3: { val: 0, fct: 50 }
-  A4: { val: 0, fct: 50 }
-  A5: { val: 0, fct: 50 }
-  A6: { val: 0, fct: 50 }
-  A7: { val: 0, fct: 50 }
-  A8: { val: 0, fct: 50 }
-  A9: { val: 0, fct: 50 }
-  A10: { val: 0, fct: 50 }
-  A11: { val: 0, fct: 50 }
-  A12: { val: 0, fct: 50 }
-  A13: { val: 0, fct: 50 }
-  A14: { val: 0, fct: 50 }
-  A15: { val: 0, fct: 50 }
+config_bonded_opti:  # must correspond to bond and angle types as defined in your input ITP files
+  B1: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]}}  # Bonds
+  B2: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }  # nm, kJ.mol-1.nm-2
+  B3: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B4: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B5: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B6: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B7: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B8: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B10: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  B11: { start: { val: 0.5, fct: 1250 }, swarm_init: { val: [0, 1], fct: [500, 2500] }, boundaries: { val: [0, 1.5], fct: [0, 5000]} }
+  A1: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }  # Angles
+  A2: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }  # degrees, kJ.mol-1
+  A3: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A4: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A5: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A6: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A7: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A8: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A9: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A10: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A11: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A12: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A13: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A14: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
+  A15: { start: { val: 150, fct: 50 }, swarm_init: { val: [120, 160], fct: [0, 50] }, boundaries: { val: [0, 180], fct: [0, 150]} }
 
 # 10. INITIAL FORCE FIELD PARAMETERS FOR NON-BONDED TERMS
 init_nonbonded:  # must correspond to bead types as defined in your input ITP files
@@ -173,18 +173,13 @@ exp_data:
 
 # 15. WEIGHT/CONFIDENCE ATTRIBUTED TO/INTO THE REFERENCE AA TRAJECTORIES
 reference_AA_weight:  # setting value to 0 will disable the usage of the bottom-up component for a given lipid
-  POPC: 1.0
-  DOPC: 1.0
-  DPPC: 1.0
-  DMPC: 1.0
-  DLPC: 1.0
-  SDPC: 0.75
-  PDPC: 0.75
-  DSPC: 0
+  POPC:
+    303K: 1.0
+  DOPC:
+    303K: 1.0
 
 # 16. OTHERS PARAMETERS (SHOULD NOT NEED CHANGES)
 sim_kill_delay: 60  # number of seconds after which a simulation will be killed if it did not write in its log file
 bonds2angles_scoring_factor: 500
-next_cycle: False  # for restarting from the single best set of FF parameters in a previous optimization output directory
 row_x_scaling: True  # use the same scale on all X axes for bond and angle distribution plots
 row_y_scaling: True  # use the same scale on all Y axes for bond and angle distribution plots

optimize_lipids.py

@@ -113,6 +113,9 @@
 ns.opti_moves_details_lipid_temp_file = 'opti_moves_details_lipid_temp.log'  # score parts and all info per lipid+temp
 ns.opti_moves_times_file = 'opti_moves_times.log'
 
+if not os.path.isdir(f"{config.data_aa_storage_dir}"):
+    os.mkdir(f"{config.data_aa_storage_dir}")
+
 # careful if continuing into an output directory of a previous optimization run
 fstpso_checkpoint_in = None  # by default we assume there is no checkpoint
 fstpso_checkpoint_in_nb = 0
@@ -942,7 +945,6 @@
         geom_grp_std = 'Unknown'
 
     # DISABLED BLOCK BELOW SO THAT WE ALWAYS START FROM ALL THE VALUES PRESENT IN THE CONFIG FILE
-    ns.params_val[geom_grp]['ref_eq_val'] = ns.user_config['init_bonded'][geom_grp]['val']
     # by default we start from the values in the config file, and we instead start from the AA average if the parameter has to be tuned
     # in case we have no AA reference data, then we fall back to the values in the config file
     # if geom_grp in ns.user_config['tune_bonds_equi_val'] and geom_grp_avg != 'Unknown':
@@ -990,7 +992,6 @@
         geom_grp_std = 'Unknown'
 
     # DISABLED BLOCK BELOW SO THAT WE ALWAYS START FROM ALL THE VALUES PRESENT IN THE CONFIG FILE
-    ns.params_val[geom_grp]['ref_eq_val'] = ns.user_config['init_bonded'][geom_grp]['val']
     # by default we start from the values in the config file, and we instead start from the AA average if the parameter has to be tuned
     # in case we have no AA reference data, then we fall back to the values in the config file
     # if geom_grp in ns.user_config['tune_angles_equi_val'] and geom_grp_avg != 'Unknown':