Biomarker Selection for Phenotype Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining and predicting phenotypes are limited by the potential infinite number of possible phenotypes for any given organism, making it challenging to improve prediction of organism performance, suggest management or health interventions, and select organisms for breeding or genetic improvement.
Innovation Solution
A method involving the selection of biomarkers, where a plurality of biomarkers are generated, and a subset is selected based on heritability and correlation with other biomarkers. Machine learning tests, such as lasso or ridge regression, are used to determine the relationship between the subset of biomarkers and the phenotype of interest, allowing for the prediction of phenotypes and optimal management strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of biomarkers are used to predict phenotypes, then prediction accuracy improves, but the complexity of the system increases
Solution Approach 1:
The patent extracts and selects only the most relevant biomarkers from a large pool of potential biomarkers. The selection process identifies biomarkers that have strong predictive power for specific phenotypes while filtering out redundant or less informative biomarkers, thus maintaining prediction accuracy while reducing system complexity.
Solution Approach 2:
The patent changes the parameters of the biomarker set by applying selection criteria based on heritability, correlation with phenotypes, and statistical significance. This parameter transformation converts a large, complex biomarker dataset into a streamlined subset that retains predictive power while reducing complexity.
2Measurement precision
If machine learning tests are applied to determine relationships between biomarkers and phenotypes, then prediction ability improves, but computational requirements increase
Solution Approach 1:
The patent applies machine learning tests selectively rather than exhaustively to all possible biomarker-phenotype combinations. By focusing computational resources on the most promising relationships identified through preliminary filtering and selection, the system achieves high prediction ability while avoiding unnecessary computational expenditure.
Solution Approach 2:
The patent performs preliminary biomarker selection and filtering before applying computationally intensive machine learning tests. This preliminary action reduces the dataset size and identifies high-priority relationships, allowing the subsequent machine learning analysis to be more efficient and less computationally demanding.
3Productivity
If heritable biomarkers are selected for breeding programs, then genetic improvement efficiency improves, but the scope of phenotypes considered may be limited
Solution Approach 1:
The patent adds the dimension of heritability estimation to the biomarker selection process. By evaluating biomarkers not only for their association with phenotypes but also for their heritability, the system enables breeding programs to focus on traits that can be effectively transmitted to offspring, thus improving breeding efficiency while maintaining versatility through multi-criteria evaluation.
Data Source
AI summary
An example method of predicting a phenotype of interest of a subject includes generating apluralityofbiomarkers; selectingasubsetofbiomarkersfromthepluralityofbiomarkers;anddeterminin garelationshipbetweenthesubsetofbiomarkersandthephenotypeofinterest;receivingsample biomarkers;where thesample biomarkerscomprise oneor more measurementsofthesubject;andpredictingthephenotypeofinterestofthesubjectbasedonthesampleb iomarkers and the relationship between the subsetof biomarkers and the phenotype of interest.