Genetic Optimization Advanced Options Dialog Box

Set options for genetic optimization of optoelectronic properties.

  • Features
  • Additional Resources

Genetic Optimization Advanced Options Dialog Box Features

Individual selection section

In this section, you specify how individuals are selected as parents for the next generation. The selection may be based on the scoring function for the "fitness" of the individuals.

Always use the top N scoring individuals without modification in the next generation text box

Specify the number of individuals to carry forward to the next generation without modification. The individuals with the highest scores are selected (elitism). Set the value to 0 to turn off this feature.

Selection method

Choose a method for selecting the individuals that are used as the parents for the next generation. The selection is done to fill the places in the population after the highest-scoring individuals are selected (if any). The choices are:

  • Rank—Select individuals at random, with the probablility of selection proportional to the rank of the individual. The rank is the index of the individual in a list sorted by increasing score: the highest scoring individual has the highest index and thus the highest rank. (This is a roulette wheel method in which the pocket size is the rank.)
  • Roulette wheel—Select individuals at random, with the probability of selection (pocket size) proportional to the score.
  • Tournament—Select the highest-scoring individual from a randomly selected pool of the size specified in the Tournament size box. The selection of members of the pool is done by uniform random sampling.
  • Tournament with roulette—Select the highest-scoring individual from a randomly selected pool of the size specified in the Tournament size box. The selection of members of the pool is done by roulette-wheel random sampling, where the probability of selection is proportional to the score.
  • Uniform—Select individuals by uniform random selections. In this method, all individuals have the same probability of selection: there is no bias toward the fitter individuals.
Tournament size option menu

Specify the number of individuals selected in each pool for the tournament selection method.

Protocol for scaling scores option menu

Choose the protocol for scaling scores to differentiate between individuals towards the end of the optimization, where scores may not be well separated. The protocols are standard genetic algorithm protocols, information on which can be found in textbooks or online.

Penalize options

Select options for penalizing certain kinds of structures. A structure score is evaluated for the structure, as described below. If the structure score is below this threshold, its optoelectronic properties are not calculated and its survival score is lowered so that it is less likely to be selected for the next generation.

  • Atypical structures—Penalize structures that are atypical for optoelectronics. The structure score is based on the number of atoms (< 200), the molecular weight (< 1100), the number of different elements (< 5), the presence of three-atom or longer chains consisting of N, O, P, or S, and the presence of two or more different halogens, with weights for each of these components in the score.
  • H2—If a genetic operation produces the hydrogen molecule, its structure score is set to -100.
Bad structure score threshold text box

Set the threshold on the structure score for determining when a structure is considered to be "bad". This score is used when introducing new structures to replace bad structures, and for determining whether to evaluate optoelectronic properties when structure-based properties are in use (either for optimization or for penalties). If you want to perform optoelectronics calculations on all structures then set this threshold to a large negative number and none will be replaced nor skipped. If you want to replace or skip a larger number of the lowest structure-scoring individuals then set this threshold to a larger positive number.

Introduce new structures to replace options

When a genetic operation produces undesirable structures, or no structures, they can be replaced by structures from a pool, which is defined by the options selected for Draw structures from. There are two classes of undesirable structures:

  • Non-unique children—Child structures from a crossover mutation that are the same as each other or the same as their parents. If the crossover produces the same results as the two parents, both children are replaced; if the children are identical to each other, one of the children is replaced. This option also replaces parents from the pool if they produce no structures.
  • Bad structures—Structures whose structure score is below the threshold given in the Bad structure score threshold text box. It is replaced by a structure from the pool that has a better structure score. This option is only relevant if you are including structure-based properties for scoring or you are penalizing structures with the Penalize options.
Draw structures from options

Specify the source of structures to draw from when replacing undesirable structures. When selecting a structure, the source is selected at random, then from the source a structure is selected at random. The available sources are:

  • Remaining input structures—Draw structures from the set of input structures that were not selected for the initial population.
  • Previous generations—Draw structures from the set of structures from all previous generations.
  • Fragments—Draw structures from the fragment libraries (see Choose Fragment Libraries). This option is not available if the input structures are monomers or polymers from the Polymer Builder Panel, as these structures have special atom properties that mark the connection points for polymerization, and the fragment libraries do not.

Property-based termination section

In this section, you can choose the criterion for terminating the optimization based on whether properties hit the target.

One property hits target option

Stop the optimization if the target for one of the properties is attained (greater than, equal to, or less than the target value). Only available if at least one property has a target value.

All properties hit targets option

Stop the optimization if the targets for all of the properties are attained (greater than, equal to, or less than the target value). Only available if more than one property has a target value.

Other options

Set random number seed option and text box

Select this option to specify a random seed to use for all random operations. Specifying the seed allows you to reproduce the results, unless other factors affect them. If this option is not selected, a seed is chosen at random.

Choose Fragment Libraries button

Click this button to choose the fragment libraries to use in fragment mutation. The available standard libraries include some of the fragment libraries available in the Build panel, and an optoelectronics library. You can also specify your own set of fragments by selecting Custom and opening a file that contains the structures. All hydrogen atoms on these structures are considered as candidates for removal to form a bond to the individual that is being mutated, and one of these is chosen at random.

Load Custom Conformational Search Options button

Use a MacroModel simplified input file to customize the conformational search. Clicking this button opens a file selector in which you can navigate to and select the file. The default MacroModel conformational search options are used otherwise. The file name is reported to the right of the button. If you no longer want to use this file, click the X button to the right of the file name.

Load Custom Optoelectronics Settings button

Load optoelectronics settings from a plain text file. Clicking this button opens a file selector in which you can navigate to and select the file. The file name is reported to the right of the button. If you no longer want to use this file, click the X button to the right of the file name.

The settings are essentially command-line options. Each Genetic Optimization job writes a file containing all the options, named optoelectronics_default_settings.txt. You can edit this file to create a custom settings file. For more information on the settings, run the following command in a terminal window:

$SCHRODINGER/run optoelectronics_gui_dir/optoelectronics_driver.py -h

Related Topics