Epistasis Feature Selection via Relevance Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computational biology, feature selection methods struggle with the 'curse of dimensionality' in gene expression and genotype data, leading to poor performance and inefficiency in modeling epistasis effects due to the large number of variables compared to samples, and existing greedy strategies often miss interactions between low-marginal effect markers.
Innovation Solution
A computer-implemented method using a feature selection module that determines relevance scores for genetic markers and their interactions, selecting a subset of top-k features based on Maximum-Relevancy and Minimum-Redundancy criteria, dynamically updating thresholds to efficiently model epistasis effects on physical traits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all genetic markers and interactions are included in the model, then the completeness of epistasis modeling is improved, but the computational efficiency deteriorates due to the curse of dimensionality
Solution Approach 1:
The patent extracts only the most relevant genetic markers and their interactions by calculating relevance scores and selecting top-k features. This extraction process removes redundant and irrelevant features from the high-dimensional dataset, retaining only those features that contribute most significantly to phenotype prediction while maintaining computational feasibility.
Solution Approach 2:
The patent segments the feature selection process into distinct stages: first selecting individual genetic markers based on relevance scores, then selecting interactions between markers. This segmentation allows the model to handle high-dimensional data in manageable steps, improving computational efficiency while systematically capturing epistasis effects.
2Measurement precision
If a large number of genetic markers are selected for modeling, then the accuracy of phenotype prediction is improved, but the model complexity increases leading to overfitting
Solution Approach 1:
The patent implements a dynamic feature selection process where the number of features to select (k) can be adjusted based on validation performance. The model dynamically balances the trade-off between model complexity and predictive accuracy by selecting an optimal subset size that prevents overfitting while maintaining sufficient predictive power.
Solution Approach 2:
The patent performs preliminary feature selection by calculating relevance scores for all genetic markers before building the final model. This preliminary action identifies and ranks important features in advance, allowing the model to focus computational resources on the most informative features and reducing the risk of overfitting by excluding irrelevant features from the outset.
3Productivity
If existing greedy strategies are used for feature selection, then the computational speed is improved, but the ability to detect interactions between low-marginal effect markers deteriorates
Solution Approach 1:
The patent introduces relevance scores as an intermediary metric that mediates between computational efficiency and interaction detection. By calculating relevance scores for all markers first, the system creates a ranked list that guides subsequent interaction analysis, enabling efficient detection of interactions between low-marginal effect markers through systematic evaluation rather than random sampling.
Solution Approach 2:
The patent implements a feedback mechanism where the selection of individual markers informs the subsequent selection of interactions. The relevance scores of individual markers serve as feedback to guide which interactions to evaluate next, allowing the algorithm to efficiently focus on interactions involving already-selected markers while maintaining the ability to detect interactions between low-marginal effect markers.
Data Source
AI summary
Various embodiments select markers for modeling epistasis effects. In one embodiment, a processor receives a set of genetic markers and a phenotype. A relevance score is determined with respect to the phenotype for each of the set of genetic markers. A threshold is set based on the relevance score of a genetic marker with a highest relevancy score. A relevance score is determined for at least one genetic marker in the set of genetic markers for at least one interaction between the at least one genetic marker and at least one other genetic marker in the set of genetic markers. The at least one interaction is added to a top-k feature set based on the relevance score of the at least one interaction satisfying the threshold.