Expert-in-the-loop AI for Polymer Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computational methods for polymer design generate a large number of candidate molecules, many of which are not synthetically viable or compliant with industrial constraints, leading to inefficiencies and bottlenecks in the polymer discovery process due to the lack of consideration for practical constraints like synthetic viability and regulatory compliance.
Innovation Solution
A computer-implemented method and system that incorporates a 'human-in-the-loop' approach, using machine learning to replicate the decisions of subject matter experts and prioritize candidates based on their input, thereby filtering out non-viable options and focusing on synthetically accessible and compliant polymer candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If computational screening and generative modeling are used to accelerate polymer design, then the number of candidate molecules generated increases, but the proportion of synthetically viable and industrially compliant candidates decreases
Solution Approach 1:
The system performs preliminary filtering of computational candidates against synthetic viability criteria and industrial constraints before they reach expert review. By pre-screening candidates based on established chemical rules and constraints (such as synthetic accessibility scores, regulatory compliance, and feedstock availability), the system eliminates non-viable candidates in advance, ensuring that only promising candidates are presented to experts for final selection.
2Adaptability or versatility
If a large number of candidates are generated computationally, then the coverage of material property space increases, but the time required for expert review increases
Solution Approach 1:
The system extracts and applies explicit synthetic viability criteria and industrial constraints as filtering rules before candidate generation. By incorporating these constraints into the computational screening process itself (rather than relying solely on post-generation filtering), the system reduces the total number of candidates requiring expert review while maintaining comprehensive coverage of viable material property space.
3Measurement precision
If manual review by subject matter experts is used to select viable candidates, then the quality of selection improves, but the bottleneck in the discovery process worsens
Solution Approach 1:
The system introduces an intelligent filtering system as an intermediary between computational candidate generation and expert review. This intermediary automatically screens candidates based on synthetic viability, industrial constraints, and chemical common sense, presenting only the most promising candidates to experts. This intermediary layer maintains high selection accuracy by preserving expert judgment while dramatically reducing the volume of candidates requiring manual review.
4Quantity of substance
If computational approaches ignore practical constraints like synthetic viability and regulatory compliance, then the number of candidates meeting target properties increases, but the gap between computational candidates and industrial realizability increases
Solution Approach 1:
The system changes the parameters used in computational screening to include explicit weights for synthetic viability, regulatory compliance, and industrial constraints. By modifying the objective function and filtering criteria to incorporate these practical considerations alongside target material properties, the system generates candidates that simultaneously optimize for both performance and manufacturability, closing the gap between computational predictions and industrial realizability.
Data Source
AI summary
A set of material candidates expected to yield materials with target properties can be generated. A subject matter expert's decision indicating accepted and rejected material candidates from the set of material candidates can be received. Based on the subject matter expert's input, a machine learning model can be trained to replicate the subject matter expert's decision.