Polymer design device, polymer design method, and non-transitory recording medium

1. A polymer design device comprising:

a processor that is configured to:

receive a requirement for a target physical property of a desired polymer;

acquire structural information of polymers; and

estimate, for each polymer corresponding to the structural information acquired by the polymer structural formula acquisition unit, physical property information of the polymer indicating a physical property of the polymer, based on the structural information of the polymer and based on a regression model representing a relationship between the structural information of the polymer and the physical property of the polymer, the physical property being a variable following a given probability distribution, and the physical property information including a mean value of the given probability distribution and a standard deviation of the given probability distribution; and

wherein the processor is further configured to:

calculate, for each polymer corresponding to the structural information acquired by the polymer structural formula acquisition unit, a score of the polymer based on the requirement for the target physical property and based on the mean value and the standard deviation of the physical property of the polymer,

rank the polymers acquired by the processor based on the score of each of the polymers, and select, from among the structural information of the polymers acquired by the processor, at least one polymer as the desired polymer, based on the score of each of the polymers, and

output information of only the selected at least one polymer, thereby reducing number of trials of finding structural information of a polymer that has the target physical property.

2. The polymer design device according to claim 1 , wherein the given probability distribution is a normal distribution.

3. The polymer design device according to claim 1 , the processor is further configured to:

calculate a feature quantity of a polymer from the structural information of the polymer; and

construct the regression model by performing a regression analysis to estimate the relationship between the feature quantity of the polymer and the physical property of the polymer.

4. The polymer design device according to claim 3 , wherein the regression analysis is Gaussian process regression, Bayesian ridge regression, ensemble regression, or deep learning.

5. The polymer design device according to claim 3 , wherein the processor is configured to calculate as the feature quantity of the polymer, from the structural information of the polymer based on a predetermined algorithm,

an Extended Connectivity Fingerprint (ECFP),

molecular weight,

log P,

a number of alicyclic rings,

a number of aromatic rings, a number of heterocyclic rings,

a quotient of the ECFP divided by a number of constituent atoms of a repeating unit of the polymer,

a quotient of the molecular weight divided by the number of the constituent atoms of the repeating unit of the polymer,

a quotient of the log P divided by the number of the constituent atoms of the repeating unit of the polymer,

a quotient of the number of alicyclic rings divided by the number of the constituent atoms of the repeating unit of the polymer,

a quotient of the number of aromatic rings divided by the number of the constituent atoms of the repeating unit of the polymer, or

a quotient of the number of heterocyclic rings divided by the number of the constituent atoms of the repeating unit of the polymer.

6. The polymer design device according to claim 1 , wherein the processor is configured to calculate the score by adding the standard deviation to the mean value, in response to receiving the requirement for the target physical property indicating maximization of a physical property.

7. The polymer design device according to claim 1 , wherein the processor is configured to calculate the score by subtracting the standard deviation from the mean value, in response to receiving the requirement for the target physical property indicating minimization of a physical property.

8. A non-transitory recording medium storing a computer program to cause a computer to perform a method comprising:

receiving a requirement for a target physical property of a desired polymer;

acquiring structural information of polymers;

estimating, for each polymer corresponding to the acquired structural information, physical property information of the polymer indicating a physical property of the polymer, based on the structural information of the polymer and based on a regression model representing a relationship between the structural information of the polymer and the physical property of the polymer, the physical property being a variable following a given probability distribution, and the physical property information including a mean value of the given probability distribution and a standard deviation of the given probability distribution;

calculating, for each polymer corresponding to the acquired structural information, a score of the polymer based on the requirement for the target physical property and based on the mean value and the standard deviation of the physical property of the polymer;

ranking the polymers acquired by the processor based on the score of each of the polymers, and selecting, from among the structural information of the polymers acquired by the acquiring, at least one polymer as the desired polymer, based on the score of each of the polymers; and

outputting information of only the selected at least one polymer, thereby reducing number of trials of finding structural information of a polymer that has the target physical property.

9. A method performed by a computer comprising:

receiving a requirement for a target physical property of a desired polymer;

acquiring structural information of polymers;

estimating, for each polymer corresponding to the acquired structural information, physical property information of the polymer indicating a physical property of the polymer, based on the structural information of the polymer and based on a regression model representing a relationship between the structural information of the polymer and the physical property of the polymer, the physical property being a variable following a given probability distribution, and the physical property information including a mean value of the given probability distribution and a standard deviation of the given probability distribution;

calculating, for each polymers corresponding to the acquired structural information, a score of the polymer based on the requirement for the target physical property and based on the mean value and the standard deviation of the physical property of the polymer;

ranking the polymers acquired by the processor based on the score of each of the polymers, and selecting, from among the structural information of the polymers acquired by the acquiring, at least one polymer as the desired polymer, based on the score of each of the polymers; and

outputting information of only the selected at least one polymer, thereby reducing number of trials of finding structural information of a polymer that has the target physical property.

10. The polymer design device according to claim 1 , wherein the processor is further configured to train the regression model using the calculated feature quantity of the polymer and the physical property of the polymer that is obtained from experiments.