Genetic Variant Detection via Machine Learning Clustering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current genome-wide association studies for autoimmune diseases face limitations such as limited power in detecting variable effect sizes, phenotypic heterogeneity, recruitment bias, and failure to identify novel associations due to rarer variants with smaller effects, as well as population stratification and genotyping platform artifacts.

Innovation Solution

The development of new genetic markers and diagnostic methods for autoimmune diseases, including pediatric autoimmune disorders, focusing on specific genetic alterations in genes like ADCY7, IL23R, PTPN22, and others, to diagnose and treat conditions like psoriasis and Crohn's disease by detecting SNVs, insertions, deletions, or copy number variations, and administering targeted pharmaceutical agents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If genome-wide association studies are conducted to identify genetic markers for autoimmune diseases, then the coverage of genetic variants is increased, but the power to detect variable effect sizes and rare variants with smaller effects is limited

Engineering Contradiction:
Improvecoverage of genetic variantsVSAvoiddetection power for rare variants
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The study segments the autoimmune disease population into distinct clusters based on genetic profiles using unsupervised machine learning algorithms. This segmentation allows for more precise identification of genetic markers within specific subgroups, improving detection power for rare variants with smaller effects that would be diluted in heterogeneous populations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention employs dynamic machine learning algorithms that can adapt and learn from data patterns to identify complex genetic interactions. This dynamic approach enables the system to detect rare variants and their interactions more effectively than static statistical methods, as the algorithms can recognize non-linear patterns and epistatic effects.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If traditional genome-wide association studies are used, then population-level genetic associations are identified, but phenotypic heterogeneity and recruitment bias limit the ability to identify novel associations

Engineering Contradiction:
Improveidentification of novel associationsVSAvoidimpact of phenotypic heterogeneity
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The study performs preliminary clustering of patients into genetically-defined subgroups before conducting association analyses. This preliminary action reduces phenotypic heterogeneity within each cluster, allowing for more reliable identification of novel genetic associations that would be obscured in heterogeneous populations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The machine learning framework incorporates feedback loops where clustering results inform subsequent association analyses, which in turn refine the clustering. This iterative feedback process continuously improves the ability to identify novel associations while accounting for phenotypic heterogeneity, as the system learns from its own results to better distinguish true signals from noise.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If genetic markers are identified for specific autoimmune disorders, then precise diagnosis and targeted treatment are enabled, but the complexity of analyzing multiple genes and variants increases

Engineering Contradiction:
Improvediagnostic precisionVSAvoidcomplexity of genetic analysis
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The invention develops a universal machine learning framework that can simultaneously analyze multiple genes, variants, and their interactions across different autoimmune diseases. This multi-functional system handles the complexity of genetic analysis by integrating multiple data types and analytical approaches into a single coherent platform, reducing the burden of separate analyses while maintaining high diagnostic precision.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The machine learning algorithms serve as intermediaries that translate complex genetic data into clinically actionable insights. These intermediary systems process the complexity of multiple genes and variants, identifying patterns and interactions that would be difficult for clinicians to interpret directly, thereby maintaining diagnostic precision while managing analytical complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3730188B1Compositions and methods for use in combination for the treatment and diagnosis of autoimmune diseases
Publication Date: 2024.07.31 THE CHILDRENS HOSPITAL OF PHILADELPHIA
  • EP3730188B1 patent drawingFigure 1A
  • EP3730188B1 patent drawingFigure 1B
  • EP3730188B1 patent drawingFigure 1C

AI summary

This disclosure provides new genetic targets, diagnostic methods, and therapeutic treatment regimens for multiple autoimmune disorders, including pediatric autoimmune disorders that are co-inherited and genetically shared. The disclosure, for example, provides methods of diagnosing or determining a susceptibility for one or more autoimmune diseases and methods of determining treatment protocols for patients with one or more autoimmune diseases based on determining if the patients have genetic alterations in particular genes.