Antisense Oligonucleotide Delivery Peptide Design with Interpretable ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing antisense oligonucleotides and peptide-oligonucleotide-conjugates face challenges in achieving improved delivery and efficacy for therapeutic applications, particularly in efficiently transporting oligonucleotides into cells and nuclei.
Innovation Solution
Development of peptide-oligonucleotide-conjugates using machine learning to identify optimal cell-penetrating peptides, enhancing delivery through conjugation with trimeric peptides optimized for cellular uptake, and employing a nested long short-term memory (LSTM) recurrent neural network model to predict peptide sequences with improved structure-activity relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional oligonucleotides are used for therapeutic applications, then gene expression modulation is achieved, but cellular delivery and nuclear uptake are insufficient
Solution Approach 1:
The patent combines oligonucleotides with cell-penetrating peptides to create conjugates that merge the gene-modifying capability of oligonucleotides with the cellular delivery capability of peptides, achieving both therapeutic efficacy and improved cellular uptake
Solution Approach 2:
Cell-penetrating peptides serve as intermediary carriers that facilitate the transport of oligonucleotides across cell membranes and into the nucleus, solving the delivery barrier without compromising the oligonucleotide's therapeutic function
2Reliability
If peptide-oligonucleotide conjugates are designed using traditional methods, then delivery is improved, but the design process is time-consuming and lacks predictive accuracy
Solution Approach 1:
The patent replaces traditional empirical trial-and-error peptide design methods with machine learning algorithms that predict optimal peptide sequences based on structural and functional criteria, dramatically reducing design time and improving predictability
Solution Approach 2:
The machine learning model creates virtual models of peptide-oligonucleotide conjugates to predict their behavior and performance before actual synthesis, allowing rapid iteration and optimization without physical trial-and-error
3Productivity
If existing cell-penetrating peptides are used, then some cellular uptake is achieved, but structure-activity relationships are insufficient for optimization
Solution Approach 1:
The patent employs machine learning models that incorporate feedback from experimental data on peptide structure and activity, continuously refining predictions to achieve precise structure-activity relationships and optimize peptide design
Solution Approach 2:
The patent systematically varies and optimizes peptide parameters such as sequence, length, and structural features using machine learning guidance, achieving precise control over cellular uptake efficiency through data-driven parameter optimization
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The conjugates demonstrate a significant increase in cellular uptake and therapeutic efficacy, with some achieving up to 40-fold improvement compared to unconjugated oligonucleotides, while maintaining low toxicity and immunogenicity.
Implementation Method 1
The conjugates demonstrate a significant increase in cellular uptake
Implementation Method 2
an antisense compound, e.g., an oligonucleotide, which hybridizes to a target nucleic acid
Data Source
AI summary
Provided herein are oligonucleotides, trimeric peptides, and peptide-oligonucleotide-conjugates. Also provided herein are methods of treating a muscle disease in a subject in need thereof, comprising administering to the subject oligonucleotides, trimeric peptides, and peptide-oligonucleotide-conjugates described herein. A synthetic method provides for the generation of a library of cell-penetrating peptides conjugated to an antisense oligonucleotide, and a machine learning-based generator-predictor-optimizer loop for the generation of novel peptide sequences capable of enhanced delivery of oligonucleotide cargo from the library of conjugates.