Generative Sequence Screening for RNA Editing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current RNA editing systems face limitations such as aberrant effector activity, delivery barriers, unintended transcriptomic modifications, and immunogenicity, and lack efficient and specific targeting of tissues like muscle and the central nervous system, while also being constrained by the complexity of regulatory elements and the vastness of potential gRNA sequences.

Innovation Solution

The use of machine learning approaches, specifically generative adversarial networks (GANs), diffusion models, and denoising diffusion conditional GANs (ddGANs), to predict and design polymer sequences for guide RNAs and AAV capsid proteins, enabling the generation of sequences with target biological properties such as high infectivity for specific tissues and low liver tropism, and improving the specificity and efficiency of RNA editing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If naturally-occurring AAV capsids are used for gene therapy delivery, then the system is simple and well-established, but the tissue targeting specificity is poor and liver tropism is high

Engineering Contradiction:

Improvetissue targeting specificityVSAvoidcapsid design complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by using machine learning models to predict and design capsid sequences with desired tissue-specific tropism properties before actual gene therapy delivery. The models pre-screen and optimize capsid variants to achieve high muscle or CNS specificity while minimizing liver tropism, thereby resolving the contradiction between targeting precision and design complexity.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent employs parameter changes by systematically varying capsid amino acid sequences and analyzing their impact on tissue tropism using machine learning. By changing specific sequence parameters and their combinations, the system identifies capsid variants with optimized tissue specificity, overcoming the limitation of naturally-occurring capsids without requiring overly complex design approaches.

Inventive Principle:

Principle #35Parameter changes

2Speed

If the dose of AAV is increased to ensure infection of desired tissues, then the infectivity of target tissues improves, but dose-dependent liver toxicity increases

Engineering Contradiction:

Improveinfectivity rateVSAvoidliver toxicity

Core Design Contradiction:

SpeedVSObject-affected harmful factors

Solution Approach 1:

The patent applies local quality by designing capsids with tissue-specific properties that concentrate infectivity in target tissues (muscle or CNS) while minimizing interaction with non-target tissues like the liver. This localized optimization of capsid-tissue affinity allows achieving high infectivity in desired tissues at lower overall doses, thereby reducing dose-dependent liver toxicity.

Inventive Principle:

Principle #3Local quality

3Productivity

If conventional methods are used to discover regulatory element sequences, then the approach is straightforward, but the search through the vast sequence space is inefficient and incomplete

Engineering Contradiction:

Improvesequence discovery efficiencyVSAvoidmodel system complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces conventional mechanical/experimental sequence discovery methods with machine learning-based predictive models. These models efficiently navigate the vast sequence space by learning from existing data and predicting functional regulatory sequences, dramatically improving discovery productivity while managing system complexity through computational approaches.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Reliability

If current RNA editing systems are used, then the basic editing function is achieved, but aberrant effector activity and unintended transcriptomic modifications occur

Engineering Contradiction:

Improveediting specificityVSAvoidsystem design complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies feedback by using machine learning models that learn from experimental outcomes to predict and optimize gRNA sequences with high editing specificity. The models incorporate feedback from training data about successful editing events and off-target effects, continuously improving predictions to reduce aberrant effector activity and unintended modifications while managing system complexity.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250014676A1Generative sequence screening with conditional gans, diffusion models, and denoising diffusion conditional gans

Publication Date: 2025.01.09 SHAPE THERAPEUTICS INC

AI summary

Systems and methods for generating a polymer sequence for a biological molecule having one or more target biological properties are provided. A plurality of target metric values for one or more target biological properties of a biological molecule and a seed for a nucleic acid or amino acid sequence for the biological molecule are inputted into a conditional generator model of a conditional generative adversarial network to obtain as output from the conditional generator model a nucleic acid or amino acid sequence for the biological molecule that is predicted by the conditional generator model to confer on the biological molecule the one or more target biological properties approximating the plurality of target metric values.