EXCOGITO is the program to investigate the mapping problem in coarse-grained modelling of biomolecules.
If you use EXCOGITO in your research please cite:
EXCOGITO, an EXtensible COarse-GraIning TOol, M Giulini, R Fiorentini, L Tubiana, R Potestio, R Menichetti, J. Chem. Inf. Model. 2024, 64, 12, 4912–4927
An Information-Theory-Based Approach for Optimal Model Reduction of Biomolecules, M Giulini, R Menichetti, MS Shell, R Potestio, Journal of chemical theory and computation 16 (11), 6795-6813
A journey through mapping space: characterising the statistical and metric properties of reduced representations of macromolecules, R Menichetti, M Giulini, R Potestio, The European Physical Journal B 94 (10), 1-26
The only requirement is to have Openmp installed on your machine.
- Install argp by using homebrew. At the terminal, run this command:
brew install argp-standalone
- Install xcode if your version is higher than MacOs 10.7. You are not required to install the Xcode App from AppStore. At the terminal, just run this command (about 15 Gb are required free on your disk even though, at the end of installation, only 2 Gb will be consumed)
xcode-select --install
In order to have access to OpenMP libraries you can install libomp by using homebrew. At the terminal, run this command:
brew install libomp
The code can be compiled using CMake. A minimal installation is obtained following these steps:
- create a directory in excogito, such as build
mkdir build
cd build- run cmake from build, calling the outer directory
cmake ..- run make
makeCmake allows to specify several options, such as the C compiler, compilation links and compilation flags. For instance, if the optimized Intel C compiler (icc) is available, step 2 may be substitued by:
cmake .. -DCMAKE_C_COMPILER=icc -DCMAKE_C_FLAGS="-Ofast -fopenmp -I./include -mkl -xSSE4.2 -parallel -ipo -mcpu=native"On MacOs, the C compiler identification should be AppleClang (check the first line printed on terminal after launching the command cmake ..).
The typical usage of the program consists in a call to excogito with one of the following options:
- optimize: to optimize the coarse-grained mapping by minimising its mapping entropy;
- random: to randomly generate coarse-grained representations and measure the associated mapping entropies;
- measure: to measure the mapping entropy of a mapping provided by the user (in the form of a .txt file);
- norm: to calculate the norm of a mapping (provided by the user) throughout a trajectory;
- cosine: to calculate pairwise distance and cosine between a pair of mappings (provided by the user) throughout a trajectory;
- distance: to calculate the distance matrix between a data set of mappings (provided by the user) over a single conformation;
- optimize_kl: to optimize the coarse-grained mapping by minimising its mapping entropy, calculated using the original Kullback-Leibler divergence;
- random_kl: to randomly generate coarse-grained representations and measure the associated mapping entropies, calculated using the original Kullback-Leibler divergence;
- measure_kl: to measure the mapping entropy of a mapping provided by the user (in the form of a .txt file), calculated using the original Kullback-Leibler divergence.
Each task can require different input files, which are provided to the program in the form of command-line options.
For further information, please type on terminal ./excogito --help or ./excogito -h
Alternatively, for printing a short usage message, please type: ./excogito --usage or ./excogito -u
After selecting which task is suitable for your purposes, read carefully the documentation below according to your choice.
The optimize task requires the protein code string and three input files: parameter, trajectory, and energy.
In order to launch the optimize task follow this syntax:
./excogito optimize -p $parameter_file.ini -t $trajectory_file.xyz -e $energy_file.txt -c $prot_code
or
./excogito optimize --p $parameter_file.ini --t $trajectory_file.xyz --e $energy_file.txt --code $prot_codeFor further information, please type on terminal ./excogito optimize
The random task requires the protein code string and three input files: parameter, trajectory, and energy.
In order to launch the random task follow this syntax:
./excogito random -p $parameter_file.ini -t $trajectory_file.xyz -e $energy_file.txt -c $prot_code
or
./excogito random --p $parameter_file.ini --t $trajectory_file.xyz --e $energy_file.txt --code $prot_codeFor further information, please type on terminal ./excogito random
The measure task requires the protein code string and four input files: parameter, trajectory, energy, and mapping.
In order to launch the measure task follow this syntax:
./excogito measure -p $parameter_file.ini -t $trajectory_file.xyz -e $energy_file.txt -c $prot_code -m $mapping_file.txt
or
./excogito measure --p $parameter_file.ini --t $trajectory_file.xyz --e $energy_file.txt --prot_code $prot_code --m1 $mapping_file.txtFor further information, please type on terminal ./excogito measure
The norm task requires the protein code string and three input files: parameter, trajectory, and mapping.
In order to launch the norm task follow this syntax:
./excogito norm -p $parameter_file.ini -t $trajectory_file.xyz -c $prot_code -m $mapping_file.txt
or
./excogito norm --p $parameter_file.ini --t $trajectory_file.xyz --prot_code $prot_code --m1 $mapping_file.txtFor further information, please type on terminal ./excogito norm
The cosine task requires the protein code string and four input files: parameter, trajectory, 1st mapping, and 2nd mapping.
In order to launch the cosine task follow this syntax:
./excogito cosine -p $parameter_file.ini -t $trajectory_file.xyz -c $prot_code -m $mapping_file.txt -n $mapping_file2.txt
or
./excogito cosine --p $parameter_file.ini --t $trajectory_file.xyz --prot_code $prot_code --m1 $mapping_file.txt --m2 $mapping_file2.txtFor further information, please type on terminal ./excogito cosine
The distance task requires the protein code string and thre input files: parameter, trajectory, mapping matrix.
In order to launch the distance task follow this syntax:
./excogito distance -p $parameter_file.ini -t $trajectory_file.xyz -c $prot_code -x $mapping_matrix_file.txt
or
./excogito distance --p $parameter_file.ini --t $trajectory_file.xyz --prot_code $prot_code --matrix $mapping_matrix_file.txt
For further information, please type on terminal ./excogito distance
The optimize_kl task requires the protein code string and three input files: parameter, trajectory, and probability.
In order to launch the optimize_kl task follow this syntax:
./excogito optimize -p $parameter_file.ini -t $trajectory_file.xyz -r $probability_file.txt -c $prot_code
or
./excogito optimize --p $parameter_file.ini --t $trajectory_file.xyz --probs $probability_file.txt --code $prot_codeFor further information, please type on terminal ./excogito optimize_kl
The random_kl task requires the protein code string and three input files: parameter, trajectory, and probability.
In order to launch the random_kl task follow this syntax:
./excogito random_kl -p $parameter_file.ini -t $trajectory_file.xyz -r $probability_file.txt -c $prot_code
or
./excogito random_kl --p $parameter_file.ini --t $trajectory_file.xyz --probs $probability_file.txt --code $prot_codeFor further information, please type on terminal ./excogito random_kl
The measure_kl task requires the protein code string and four input files: parameter, trajectory, probability, and mapping.
In order to launch the measure_kl task follow this syntax:
./excogito measure_kl -p $parameter_file.ini -t $trajectory_file.xyz -r $probability_file.txt -c $prot_code -m $mapping_file.txt
or
./excogito measure_kl --p $parameter_file.ini --t $trajectory_file.xyz --probs $probability_file.txt --prot_code $prot_code --m1 $mapping_file.txtFor further information, please type on terminal ./excogito measure_kl
The optimize_spins task requires the protein code string and three input files: parameter, trajectory, and probability.
In order to launch the optimize_spins task follow this syntax:
./excogito optimize_spins -p $parameter_file.ini -t $trajectory_file.xyz -r $probability_file.txt -c $prot_code
or
./excogito measure_kl --p $parameter_file.ini --t $trajectory_file.xyz --probs $probability_file.txt --prot_code $prot_codeHere the trajectory is simply given by the states of a discrete system (like this).
For further information, please type on terminal ./excogito optimize_spins.
As shown in Section 2.x, the protein code string and two files are always mandatory, namely the parameter file and the xyz trajectory file. The other files can be mandatory, depending on the chosen task.
What are these files?
- $parameter_file.ini → Set of parameters in ini format for the algorithm (see 3.1). Examples are present in /examples/parameters;
- $trajectory_file.xyz → Trajectory in xyz format (see the Section 3.2). An example is present in /examples/trajectories;
- $energy_file.txt → File with the energies corresponding to each configuration in the trajectory (see the Section 3.3). An example is present in /examples/energies;
- $prot_code → Unique string that identifies the structure (see 3.4). It will be used to generate the output files;
- $mapping_file.txt → Mapping file, containing the indices of the retained atoms (see 3.5). An example is present in /examples/mappings;
- $mapping_file2.txt → 2nd Mapping file, containing the indices of the retained atoms (see 3.5). An example is present in /examples/mappings;
- $mapping_matrix_file.txt → Matrix with n_mappings CG mappings (see 3.6).
- $probability_file.txt → File with the probabilities corresponding to each configuration in the trajectory (see 3.7). They must sum to 1.0. An example is present in /examples/probabilities;
The core element of EXCOGITO is the parameter file, which is employed to define the constants used in the different tasks.
A sample parameter file for each task can be found in /examples/parameters.
There exist 16 parameters, but only few of them are mandatory for the selected task. They are illustrated in the following table:
| Parameter | Description | Type | Mandatory | Suggested value |
|---|---|---|---|---|
| atomnum | number of atoms in the system | int | all | |
| frames | number of frames in the trajectory | int | all | < 5000 on laptops, < 15000 if criterion != 3 |
| cgnum | number of CG sites | int | all | between atomnum/20 and atomnum/2 |
| criterion | criterion for clustering | int | O-R-M | 0, 1, 2, 3, or 4 |
| nclust | number of CG macrostates | int | C0 - C3 | between frames/500, and frames/100 |
| n_mappings | number of mappings in tasks random and distance | int | R-D | |
| MC_steps | number of MC step in task optimize | int | O | > 5000 |
| rotmats_period | MC steps between two full alignments in task optimize | int | O | |
| t_zero | starting temperature in task optimize | double | O | |
| distance | cophenetic distance threshold | double | C1 | |
| max_nclust | upper number of clusters | int | C2 | between frames/100 and frames/50 |
| min_nclust | lower number of clusters | int | C2 | between frames/1000 and frames/500 (must be < max_nclust) |
| Ncores | number of cores | int | no | |
| decay_time | governs temperature decay in task optimize | double | O | |
| rsd | use rsd (if 1) instead of rmsd (if 0) | int | no | |
| stride | number of structures between two pivot configurations | int | C3 | ~ 10 if frames between 1000 and 10000 |
O-R-M-D refer to the tasks (optimize/optimize_kl, random/random_kl, measure/measure_kl, distance) in which the parameter is mandatory. C0 .. C3 indicates that the parameter is mandatory if the clustering criterion is equal to 0 .. 3, respectively.
Four criteria for hierarchical clustering:
- 0 Maxclust clustering: configurations are lumped into Nclust macrostates;
- 1 Maxdist clustering: clustering with the cophenetic distance;
- 2 Multiple maxclust: as described in Giulini et al. (JCTC, 2020);
- 3 Fast clustering: as in criterion 0, but applied to a set of pivot configurations. Labels of intermediate structures are assigned to the closer pivot;
The trajectory should be provided in the xyz format. The first line of each frame indicates the number of atoms, while the second can contain an arbitrary string. As an example, a trajectory with 2 frames and 3 atoms should resemble the following string:
3
X 2.53 2.09 3.55
X 2.57 1.95 3.51
X 2.45 1.87 3.46
3
X 2.69 1.96 3.40
X 2.80 1.91 3.43
X 2.67 2.03 3.28
In the python subdirectory there is a script that helps with the conversion from GROMACS XTC to the XYZ format.
Energy files, mandatory for tasks optimize, random, and measure, should contain one value for each frame in the trajectory.
The protein code is a string that is used to create output files. Don't insert spaces or special characters in this string
A mapping file, mandatory for tasks measure, norm, and cosine is a file with an integer per line. The value correspond to the index of the atom in the xyz trajectory. As an example, a mapping with 8 sites on a peptide of 50 sites should respect the following format:
3
7
19
21
26
34
40
47
A mapping matrix is mandatory for task distance. It is simply a series of transposed mappings. If we aim at computing the distance matrix between three mappings with 8 sites on a peptide of 50 sites, we must respect the following syntax:
3 7 19 21 26 34 40 47
2 8 19 24 25 38 41 44
0 10 12 20 29 31 35 49
Probability files, mandatory for tasks optimize_kl, random_kl, and measure_kl, must contain one value for each frame in the trajectory and should be properly normalized to 1. For a trajectory of 5 frames, the following file is acceptable:
0.1
0.15
0.6
0.05
0.1
Inside the directory examples there are example files for the 6d93 protein, allowing the user to try all the different tasks:
-
optimize:
./build/excogito optimize -p examples/parameters/parameters_optimize_6d93_N31_small.ini -t examples/trajectories/6d93_100frames.xyz -e examples/energies/6d93_energies_100frames.txt -c 6d93 -
random:
./build/excogito random -p examples/parameters/parameters_random_6d93_N31_small.ini -t examples/trajectories/6d93_100frames.xyz -e examples/energies/6d93_energies_100frames.txt -c 6d93 -
measure:
./build/excogito measure -p examples/parameters/parameters_loadca_6d93_N31.ini -t examples/trajectories/6d93_1000frames.xyz -e examples/energies/6d93_energies_1000frames.txt -c 6d93 -m examples/mappings/tamapin_ca_mapping.txt -
norm:
./build/excogito norm -p examples/parameters/parameters_norm_6d93_N31.ini -t examples/trajectories/6d93_1000frames.xyz -e examples/energies/6d93_energies_1000frames.txt -c 6d93 -m examples/mappings/tamapin_ca_mapping.txt -
cosine:
./build/excogito cosine -p ./examples/parameters/parameters_cosine_6d93_N31.ini -t ./examples/trajectories/6d93_1000frames.xyz -e ./examples/energies/6d93_energies_1000frames.txt -c 6d93 -m ./examples/mappings/tamapin_ca_mapping.txt --m2 ./examples/mappings/tamapin_nextca_mapping.txt -
distance:
./build/excogito distance -p examples/parameters/parameters_distance_6d93_N31.ini -t ./examples/trajectories/6d93_1frame.xyz -x examples/mappings/6d93_mapping_matrix.txt -c 6d93 -
optimize:
./build/excogito optimize_kl -p examples/parameters/parameters_optimizekl_6d93_N31_notemp.ini -t examples/trajectories/6d93_100frames.xyz -r examples/probabilities/6d93_probs_100frames.txt -c 6d93 -
random_kl:
./build/excogito random_kl -p examples/parameters/parameters_randomkl_6d93_N31.ini -t examples/trajectories/6d93_100frames.xyz -r examples/probabilities/6d93_probs_100frames.txt -c 6d93 -
measure_kl:
./build/excogito measure_kl -p examples/parameters/parameters_measurekl_6d93_N31.ini -t examples/trajectories/6d93_100frames.xyz -r examples/probabilities/6d93_probs_100frames.txt -c 6d93 -m examples/mappings/tamapin_ca_mapping.txt -
optimize_spins
./build/excogito optimize_spins -p examples/parameters/parameters_spins_m1.ini -t examples/trajectories/m1_spins_grouped.csv -r examples/probabilities/m1_probs.txt -c m1
The approximated mapping entropy is calculated (tasks optimize, random and measure) without the scaling factor (see. Giulini et al.). This factor should be computed by the user according to the temperature employed to simulate the system.
File refman.pdf in the docs directory contains detailed documentation authomatically generated with doxygen version 1.8.5.
A custom documentation can be generated in html and tex format by running doxygen excogito_doxygen.conf.
Marco Giulini ([email protected]) Raffaello Potestio ([email protected])
