Molecular Structure Generation via Clustered Evolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional molecular structure generation methods tend to produce a large number of similar structures once a desired property value is satisfied, limiting diversity and making it difficult to synthesize, obtain raw materials, or manufacture these structures due to practical constraints such as synthesis methods, availability of materials, or high costs.
Innovation Solution
A method involving the classification of initial molecules into clusters based on feature amounts, selecting starting molecules with maximum confidence limits, and repeatedly evolving these molecules through evolutionary development steps to generate new structures, ensuring diversity and satisfaction of desired property values without localization around specific structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional molecular structure generation methods are used to satisfy desired property values, then the desired properties are achieved, but the generated structures become similar and lack diversity
Solution Approach 1:
The patent segments the molecular structure generation process into multiple independent search processes, each starting from different initial molecules. This segmentation allows parallel exploration of different structural regions in the chemical space, preventing convergence to similar structures while maintaining property value satisfaction through individual optimization of each search process.
2Reliability
If molecular structure generation focuses on satisfying desired property values, then property requirements are met, but synthesis feasibility and manufacturing practicality are compromised
Solution Approach 1:
The patent performs preliminary action by pre-defining multiple initial molecules with diverse structures before the generation process begins. These initial molecules are carefully selected to represent different structural regions and synthesis feasibility profiles. By establishing this diverse foundation in advance, the subsequent generation process can explore multiple viable synthesis pathways while maintaining property value satisfaction, avoiding the trap of generating only theoretically optimal but practically infeasible structures.
3Reliability
If a single molecular structure is optimized for desired properties, then property values are satisfied, but the solution becomes localized and lacks exploration of alternative structures
Solution Approach 1:
The patent merges multiple independent search processes into a unified generation system. Each search process independently optimizes for property values starting from different initial molecules, and the results are combined to produce a diverse set of candidate structures. This merging approach maintains the benefits of focused optimization in each search process while achieving overall solution diversity through the aggregation of multiple search outcomes.
Data Source
AI summary
To provide a molecular structure generation method and a non-transitory computer-readable medium storing a program capable of generating various molecular structures while satisfying desired property values so as not to be localized around a specific molecular structure.A molecular structure generation method according to the present invention includes: a selection step of classifying a plurality of initial molecules prepared in advance into clusters based on a feature amount and selecting a starting molecule having a maximum confidence limit value from each of the classified clusters. The method further includes an evolutionary development step of evolving each of the starting molecules. Further the selection step and the evolutionary development step are repeatedly executed for all molecules including the initial molecules and the evolved starting molecules to generate a new molecular structure.