Molecule Generation by Latent Manifold Traversal and Grammar VAE

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing computational models for drug discovery struggle with intelligent exploration of latent spaces, leading to random and invalid generation of molecular structures, which is time-consuming and costly.

Innovation Solution

A system utilizing a grammar variational auto-encoder (VAE) with manifold traversal and heuristic search to explore latent chemical spaces, incorporating expert-defined heuristics like QED and synthetic accessibility, to generate and optimize molecular structures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If existing computational models are used for drug discovery, then molecular structures can be generated, but the generation is random and invalid, leading to time-consuming and costly exploration

Engineering Contradiction:
Improvevalidity of molecular structure generationVSAvoidtime for exploring latent space
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training a grammar VAE model on historical drug data to learn the latent space structure and molecular grammar rules before actual molecule generation. This pre-processing of the latent space and establishment of grammar constraints enables more reliable and valid molecular structure generation without random exploration, directly addressing the contradiction between reliability and time loss.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If existing computational models generate molecular structures, then molecules can be produced, but the structures are often invalid and require extensive validation

Engineering Contradiction:
Improveefficiency of molecular structure generationVSAvoidvalidity of generated molecules
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements feedback by using a grammar VAE model that incorporates grammar rules as constraints during the generation process. The model continuously references the learned grammar structure and molecular rules, providing feedback mechanisms that guide the generation toward valid molecular structures. This feedback loop ensures both high productivity and reliability by preventing invalid structures rather than requiring extensive post-generation validation.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If random exploration of latent space is performed, then molecular structures can be generated, but the process is inefficient and generates invalid structures

Engineering Contradiction:
Improveability to generate diverse molecular structuresVSAvoidefficiency of molecule generation
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent applies parameter changes by transforming the latent space representation and incorporating grammar rule constraints as additional parameters during generation. Instead of random exploration, the model uses controlled parameter adjustments guided by the learned grammar structure and molecular rules. This approach maintains adaptability to generate diverse valid molecules while significantly improving productivity by eliminating random无效 exploration.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3869513B1De novo generation of molecules using manifold traversal
Publication Date: 2026.01.28 ACCENTURE GLOBAL SOLUTIONS LTD
  • EP3869513B1 patent drawingFigure 1
  • EP3869513B1 patent drawingFigure 2
  • EP3869513B1 patent drawingFigure 3

AI summary

The present disclosure relates to systems, methods, and products for identifying candidate molecule. The system includes a non-transitory memory storing instructions; and a processor in communication with the non-transitory memory. The processor executes the instructions to cause the system to receive drug data; convert the drug data into at least one point in a latent space using a grammar variational auto-encoder (VAE) model; receive a query for the at least one candidate molecule; select one or more points in the latent space; and create a k-dimensional tree graph based on the query for the at least one candidate molecule and the selected one or more points; determine a plurality of paths according to an interpolation technique; receive preference data; determine an optimum path; determine at least one candidate point on the optimum path; and determine a drug molecular structure using an inverse of the grammar VAE model.