Molecule Design Model for Property-Guided Therapeutic Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for enhancing molecular properties, particularly in small and large molecule therapeutics, face challenges in efficiently improving properties such as binding affinity, specificity, and developability, due to the complexity and size differences between these molecules, which affect their formulation and delivery.

Innovation Solution

A machine learning-based technique using a molecule design computation model that encodes and decodes molecular representations to generate output molecules with enhanced properties by training on matched datasets of molecule pairs with different property values, applying autoencoders and denoisers to guide the generation of molecules with superior properties.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional molecular design methods are used to enhance molecular properties, then the process is straightforward and easy to understand, but the efficiency and effectiveness of improving properties such as binding affinity, specificity, and developability are insufficient

Engineering Contradiction:

Improveefficiency of enhancing molecular propertiesVSAvoidcomplexity of molecular design system

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces a machine learning model as an intermediary between the molecular structure and its properties. The model takes molecular inputs (SMILES strings, graphs, or sequences) and predicts enhanced molecular properties, acting as a mediator that bridges traditional molecular design and property optimization without requiring direct complex simulations

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system optimizes molecular parameters such as binding affinity, specificity, and developability by generating modified molecular structures. The machine learning model predicts how changes in molecular parameters (composition, conformation, functional groups) affect target properties, enabling efficient parameter optimization

Inventive Principle:

Principle #35Parameter changes

2Reliability

If molecular weight is increased to create large molecule therapeutics, then the ability to modulate biochemical processes is enhanced, but the ease of oral administration and formulation is reduced

Engineering Contradiction:

Improveeffectiveness of biochemical modulationVSAvoidease of oral administration

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The patent applies local quality modifications to molecules by selectively adding or modifying specific functional groups, domains, or molecular features at particular locations. This allows enhancement of biochemical modulation capability in specific regions of the molecule while maintaining other regions that facilitate oral administration and formulation

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system segments molecular design into separate optimizable components, allowing independent optimization of therapeutic effectiveness features versus formulation and delivery features. This enables creating large molecules with enhanced biochemical activity while separately optimizing properties for oral administration

Inventive Principle:

Principle #1Segmentation

3Manufacturing precision

If molecular structure is highly optimized for specific therapeutic properties, then the binding affinity and specificity are improved, but the complexity of synthesis and manufacturing increases

Engineering Contradiction:

Improveprecision of binding affinity and specificityVSAvoidease of chemical synthesis

Core Design Contradiction:

Manufacturing precisionVSEase of manufacture

Solution Approach 1:

The machine learning model performs preliminary prediction and optimization of molecular structures before actual synthesis. By pre-screening and optimizing molecular designs computationally, the system identifies structures with high binding affinity and specificity that are also more amenable to synthesis, reducing the complexity of actual manufacturing

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250364089A1Machine learning enabled enhancement of molecular properties

Publication Date: 2025.11.27 F HOFFMANN LA ROCHE INC

AI summary

An input molecule exhibiting a value for one or more properties may be identified. A molecule design computation model may be applied to generate one or more output molecule exhibiting a different value for the one or more properties than the input molecule. The molecule design computation model may generate the one or more output molecules by at least encoding the input molecule to generate an embedding of the input molecule, and decoding the embedding of the input molecule to generate the one or more output molecules. In some cases, the molecule design computation model may generate the one or more output molecules by denoising an input molecule while conditioned on the input molecule. In some cases, the molecule design computation model may operate on a joint representation of the input molecule that combines a linear and a three-dimensional representation of the input molecule.