Genotype Labeling via Hierarchical Hidden Markov Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assigning labels to human genotypes, such as those indicating ancestry, are inefficient and inaccurate, particularly for admixed genotypes, as they fail to correctly associate alleles with haplotypes and do not update models effectively based on assigned labels.
Innovation Solution
The system employs hidden Markov models, specifically diploid and inter-window HMMs, to accurately assign labels by comparing SNP sequences from an input sample genome to reference genomes, using diploid HMMs to build label pair probability distributions and inter-window HMMs to determine label assignments, which can be updated based on assigned labels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional labeling methods are used to assign labels to genotypes, then the process is simpler, but the accuracy is poor especially for admixed genotypes
Solution Approach 1:
The patent segments the genotype data into multiple windows along the genome and processes each window independently through separate HMMs. This segmentation allows the complex labeling problem to be divided into manageable parts while maintaining overall accuracy, especially for admixed genotypes where different regions may have different ancestral origins.
Solution Approach 2:
The patent introduces a hierarchical dimension by building diploid HMMs from haploid HMMs, and then constructing an inter-window HMM that integrates results across multiple windows. This multi-level hierarchical structure enables accurate label assignment by combining information from different genomic regions and haplotype phases.
2Measurement precision
If models are updated continuously based on assigned labels, then the accuracy improves over time, but the computational resources and time required increase
Solution Approach 1:
The patent performs preliminary model training using reference panel data before actual label assignment. The diploid and inter-window HMMs are pre-built and stored, allowing rapid label assignment without repeated model training. This preliminary action separates the computationally intensive model building phase from the efficient label assignment phase.
Solution Approach 2:
The patent implements a feedback mechanism where assigned labels from new samples are used to update the reference panel and retrain the HMMs. This continuous feedback loop improves model accuracy over time by incorporating real-world data, while the incremental update approach minimizes computational overhead compared to complete retraining.
3Measurement precision
If diploid HMMs are built from haploid HMMs and integrated through inter-window HMM, then the label assignment becomes more accurate, but the computational complexity increases
Solution Approach 1:
The patent segments the HMM construction process into distinct hierarchical levels: haploid HMMs for individual chromosomes, diploid HMMs combining both haplotypes, and inter-window HMMs integrating multiple genomic regions. This segmentation makes the complex model construction manageable and computationally efficient at each stage.
Solution Approach 2:
The patent merges results from multiple haploid HMMs into diploid HMMs, and then merges diploid HMM results across windows into the inter-window HMM. This progressive merging strategy combines information from different sources systematically, improving accuracy while maintaining computational efficiency through modular integration.
Data Source
AI summary
An input sample SNP genotype is divided into a plurality of windows, each including a sequence of SNPs. For each window, a diploid hidden Markov Model (HMM) is built and from a haplotype Markov Model (MM). The diploid HMM for a window is used to determine the probability that the window corresponds to a pair of labels (e.g., ethnicity labels). An inter-window HMM, with a set of states for each window, is built based on the diploid HMMs for each window. Labels are assigned to the input sample genotype based on the inter-window HMM.