In computer-aided molecular design (CAMD) [1][2], the capability of generating new molecular species from existing one is vital. MARS program [3], devised for such demand, consists of two components:
- Molecular data structure (MDS) : base elements and 5 arrays of integers
- Genetic operators : ring formation, addition, subtraction, exchange, crossover, and combination
To initiate a MARS task, one should input the 3D structures of starting molecules. These structures will then be converted into MDS representation, where a structure is recognized as a network of base elements. Each of the genetic operators will be applied to each of the possible substructures in each of the MDSs. As a result, a number of new species can be generated.
MARS+ is based on MARS [3], with various improvements:
-
The expansion of base element library
1-1. Group-like elements are allowed. 1-2. Cationic cores and anionic cores are included. 1-3. The library is not hard-coded. It can be appended easily. (See the instructions in inputs/element_lists/element_list.txt)
-
The generalization of MDS
2-1. An additional array of integers is used to bookkeep optical isomerisms. 2-2. Two additional arrays of integers are used to bookkeep cis-trans isomerisms. 2-3. An additional array of integers is used to bookkeep cyclic bonds. 2-4. Multiple ring numbers on an atom are allowed. 2-5. The representation of a two-component chemical is allowed. Equimolar ionic liquids (1:1) are taken as examples.
-
The generalization of genetic operators
3-1. Refinement of old operators: 3-1-1. The feasibility of molecular connectivity is ensured after subtraction operation. 3-1-2. Multiple ring numbers on an atom can be induced through cyclization operator. 3-1-3. For most of the operators, the codes are revised for better systematicness. Unnecessary dependencies among variables are reduced. 3-1-4. The stability and flexibility of operators are enhanced. 3-2. Development of new operators: insertion, decyclization, element change, cis-trans inversion, chirality inversion, and component switch. 3-3. Since the new operator scheme has a better reversibility, undoing an operation is much easier now. This can reduce unintended biases toward the generation of certain molecules (e.g. polycyclics). 3-4. Check isomerisms of substructures after an operation is successfully conducted. Assign default isomerisms (i.e. trans and clockwise winding) to potentially isomeric substructures. 3-5. If an imine-like substructure is present, indicate the lone pair of nitrogen atom by a null atom "*".
-
Wrapping some Open Babel [4] functions into MARS+
4-1. This facilitates the inputting of starting structures to the program. One only needs to input SMILES now. 4-2. The isomerism perception for inputted molecules is more robust.
- Linux CentOS 6.3 (x86_64)
- Miniconda3 v4.9.2 (x86_64, py39)
- GNU g++ compiler v9.2.0 (C++11)
- Open Babel v3.1.0
- Eigen v3.3.7
- Cmake v3.15.5
- Make v4.2
The MARS+ source code consists of 7 header files and 7 cpp files: (see src/
directory)
ELEMENTS.h MOLECULE.h CASES_NEU.h CASES_IL_INDEPENDENT.h CASES_IL.h UTILITY.h PARAMETER.h
ELEMENTS.cpp MOLECULE.cpp CASES_NEU.cpp CASES_IL_INDEPENDENT.cpp CASES_IL.cpp UTILITY.cpp main.cpp
Before compiling MARS+, relevant softwares should be installed.
With Miniconda or Anaconda,
one can directly import MARS+
environment along with the required packages from MARS+_env.yml
or MARS+_env_pipcmake.yml
.
(Note: conda
may lead to a corrupted installation of cmake
.
Therefore, we provided an alternative MARS+_env_pipcmake.yml
, which installs cmake
from pip
.)
conda env create --file MARS+_env_pipcmake.yml
Now activate MARS+
environment and compile the source code.
conda activate MARS+
cd src/
rm -r ./Makefile ./cmake_install.cmake ./CMakeCache.txt ./CMakeFiles 2> /dev/null
cmake ./CMakeLists.txt
make -j [N] ("-j [N]" is optional. It means parallel compiling with N jobs at once.)
An executable MARS-PLUS
will be generated in src/
.
Make sure the compiling is successful before launching MARS+ tasks, especially when you have made modifications to the code.
Please refer to the repository of MARS+ Windows version.
There are 3 input files for MARS+: (see inputs/
)
inputs/control.in : controls the input, output, and calculation options.
inputs/element_lists/element_list.txt : a list that defines base element library.
inputs/INPUT_CHEMICALS/IL4.txt : the starting chemicals.
Please read the instructions in inputs/control.in
and inputs/element_lists/element_list.txt
.
Make sure you have properly set the parameters, and then launch the MARS+.
cd src/
./MARS-PLUS ../inputs/control.in
Alternatively, you may use the PBS scheduler. A PBS template src/job.sh
is provided.
cd src/
qsub ./job.sh
The results for each of the operations will be outputted to logs/
.
For instance, the generated chemicals by applying bond change operation to an IL will be written to logs/change_bnd_IL.txt
.
Please refer to the repository of MARS+ Windows version.
Chen-Hsuan Huang (f07524028@ntu.edu.tw) and Shiang-Tai Lin (stlin@ntu.edu.tw).
Department of Chemical Engineering, National Taiwan University, Taipei, Taiwan
[1] Austin, N. D.; Sahinidis, N. V.; Trahan, D. W., Computer-aided molecular design: An introduction and review of tools, applications, and solution techniques. Chem. Eng. Res. Des. 2016, 116, 2-26.
[2] Hsu, H. H.; Huang, C. H.; Lin, S. T., Fully Automated Molecular Design with Atomic Resolution for Desired Thermophysical Properties. Ind. Eng. Chem. Res. 2018, 57, (29), 9683-9692.
[3] Hsu, H.-H.; Huang, C.-H.; Lin, S.-T., New Data Structure for Computational Molecular Design with Atomic or Fragment Resolution. J. Chem. Inf. Model. 2019, 59, (9), 3703-3713. (https://github.com/hsuhsuanhao/MARS)
[4] O’Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.; Hutchison, G. R., Open Babel: An open chemical toolbox. J. Cheminf. 2011. (https://github.com/openbabel/openbabel)
- Neutral elements
ID | Name | Bond order | charge |
---|---|---|---|
1 | [CH0](-)(-)(-)(-) | 1 1 1 1 | 0 |
2 | C(=)(-)(-) | 2 1 1 | 0 |
3 | C(#)(-) | 3 1 | 0 |
4 | C(=)(=) | 2 2 | 0 |
5 | O(-)(-) | 1 1 | 0 |
6 | O(=) | 2 | 0 |
7 | N(-)(-)(-) | 1 1 1 | 0 |
8 | N(=)(-) | 2 1 | 0 |
9 | N(#) | 3 | 0 |
10 | O(-) | 1 | 0 |
11 | F(-) | 1 | 0 |
12 | Cl(-) | 1 | 0 |
13 | Br(-) | 1 | 0 |
14 | I(-) | 1 | 0 |
19 | S(-)(-) | 1 1 | 0 |
20 | S(=) | 2 | 0 |
21 | P(-)(-)(-) | 1 1 1 | 0 |
22 | P(=)(-) | 2 1 | 0 |
23 | P(#) | 3 | 0 |
31 | [PH0](-)(-)(-)(-)(-) | 1 1 1 1 1 | 0 |
32 | [PH0](=)(-)(-)(-) | 2 1 1 1 | 0 |
34 | S(=)(-)(-) | 1 1 | 0 |
61 | [SH0](=)(=)(-)(-) | 2 2 1 1 | 0 |
62 | Cl(=)(=)(=)(-) | 2 2 2 1 | 0 |
66 | P(=)(-)(-) | 2 1 1 | 0 |
67 | [CH0@@](-)(-)(-)(-) | 1 1 1 1 | 0 |
68 | [CH0@](-)(-)(-)(-) | 1 1 1 1 | 0 |
69 | [PH0](-)(-)(-)(-) | 1 1 1 1 | 0 |
70 | *(-) | 1 | 0 |
- Cationic elements
ID | Name | Bond order | charge |
---|---|---|---|
15 | [NH0+](-)(-)(-)(-) | 1 1 1 1 | 1 |
16 | [NH0+](=)(-)(-) | 2 1 1 | 1 |
17 | [PH0+](-)(-)(-)(-) | 1 1 1 1 | 1 |
18 | [PH0+](=)(-)(-) | 2 1 1 | 1 |
36 | [CH0](-)(-)(-)([N+]1C=CN(C)C=1) | 1 1 1 | 1 |
37 | [CH0](-)(-)(-)([N+]1C=CN(C)C(C)=1) | 1 1 1 | 1 |
38 | [CH0](-)(-)(-)(N1C=C[N+](C)=C1) | 1 1 1 | 1 |
39 | [CH0](-)(-)(-)([N+]1=CC=CC(C)=C1) | 1 1 1 | 1 |
40 | [CH0](-)(-)(-)(C1=C[N+](C)=CC=C1) | 1 1 1 | 1 |
41 | C(-)(1=[NH+]C=CC=C1) | 1 | 1 |
42 | [NH0+](-)(1=CC=CC=C1) | 1 | 1 |
57 | [In+3](-)(-)(-)(-) | 1 1 1 1 | 3 |
64 | [Ga+3](-)(-)(-)(-) | 1 1 1 1 | 3 |
65 | [SH0+](-)(-)(-) | 1 1 1 | 1 |
- Anionic elements
ID | Name | Bond order | charge |
---|---|---|---|
24 | [F-] | 0 | -1 |
25 | [Cl-] | 0 | -1 |
26 | [Br-] | 0 | -1 |
27 | [I-] | 0 | -1 |
28 | [OH1-] | 0 | -1 |
29 | [OH0-](-) | 1 | -1 |
30 | [PH0-](-)(-)(-)(-)(-)(-) | 1 1 1 1 1 1 | -1 |
33 | S(-)(=O)(=O)([O-]) | 1 | -1 |
35 | [NH0-](-)(-) | 1 1 | -1 |
43 | [NH0-](S(=O)(=O)C(F)(F)(F))(S(=O)(=O)C(F)(F)(F)) | 0 | -1 |
44 | [BH0-](-)(-)(-)(-) | 1 1 1 1 | -1 |
45 | C(-)(-)(-)(C(=O)([O-])) | 1 1 1 | -1 |
46 | C(#N)([S-]) | 0 | -1 |
47 | C(-)(-)(-)(OP(=O)(OC)([O-])) | 1 1 1 | -1 |
48 | C(#N)([N-]C#N) | 0 | -1 |
49 | [BH0-](C#N)(C#N)(C#N)(C#N) | 0 | -1 |
50 | S(OCCOCCOC)(=O)(=O)([O-]) | 0 | -1 |
51 | S(c(cc1)ccc1C)(=O)(=O)([O-]) | 0 | -1 |
52 | [PH0-](F)(F)(F)(C(C(F)(F)F)(F)F)(C(C(F)(F)F)(F)F)(C(C(F)(F)F)(F)F) | 0 | -1 |
53 | [NH0-](S(=O)(=O)C(C(F)(F)F)(F)F)(S(=O)(=O)C(C(F)(F)F)(F)F) | 0 | -1 |
54 | [CH0-](S(C(F)(F)(F))(=O)(=O))(S(C(F)(F)(F))(=O)(=O))(S(C(F)(F)(F))(=O)(=O)) | 0 | -1 |
55 | [PH0-](F)(F)(F)(F)(F)(F) | 0 | -1 |
56 | [In+3]([Cl-])([Cl-])([Cl-])([Cl-]) | 0 | -1 |
58 | Cl(=O)(=O)([O-])(=O) | 0 | -1 |
59 | [CH0-](-)(-)(-) | 1 1 1 | -1 |
60 | [NH0+](=O)([O-])([O-]) | 0 | -1 |
63 | [SH0-](-) | 1 | -1 |
- Uni-molecular operations - addition, insertion, subtraction, element change, bond change, cyclization, decyclization, cis-trans inversion, and chirality inversion:
- Bi-molecular operations - crossover:
- Bi-molecular operations - combination:
- Bi-supermolecular operation - component switch:
MARS+ supports the iterative molecular. It can be done by subjecting the molecules generated in a round to next round of exhaustive operations. A dataset obtained in this way is provided in IterDesignData repository.