De Bruijn Graph Variant Calling for HLA Typing Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for variant calling, particularly in high-throughput sequencing data, face challenges in accurately aligning highly related genomic sequences and identifying larger changes, especially in low-complexity regions and for HLA typing, with existing tools showing less than desirable accuracy and power for detecting heterozygous and homozygous sites.
Innovation Solution
A method using a de Bruijn graph approach with a reference sequence and patient sequence reads decomposed into k-mers, where each patient sequence read provides a weighted vote for HLA alleles, and an adjusted composite match score is used to rank and re-rank alleles, reducing the weight of k-mers matching the top-ranking allele to prioritize genetically less related alleles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional alignment methods are used for variant detection, then processing speed is improved, but accuracy in aligning highly related genomic sequences deteriorates
Solution Approach 1:
The patent segments the genomic sequence analysis into two distinct phases: (1) initial rapid alignment using traditional methods to locate candidate regions, and (2) refined local realignment using de Bruijn graphs specifically for highly related sequences. This segmentation allows each method to operate in its optimal performance zone, maintaining speed while improving accuracy where needed.
Solution Approach 2:
The patent applies different quality standards and methods to different regions of the genome. For most regions, traditional alignment suffices, but for regions containing highly related sequences (such as HLA genes), the patent implements local de Bruijn graph-based realignment with higher accuracy parameters, ensuring that computational resources are focused where they are most needed.
2Measurement precision
If local realignment and assembly are performed to improve variant detection accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary filtering to identify candidate regions containing highly related sequences before applying the computationally intensive de Bruijn graph realignment. By pre-identifying target regions through initial alignment and variant scoring, the patent avoids constructing de Bruijn graphs for the entire genome, thus reducing overall computational complexity while maintaining high accuracy where required.
Solution Approach 2:
The patent applies the sophisticated de Bruijn graph realignment method selectively to specific genomic regions where it is most beneficial (regions with high sequence similarity and potential variants), rather than applying it uniformly across the entire genome. This partial application strategy achieves high variant detection accuracy in critical regions while keeping overall computational complexity manageable.
3Reliability
If probabilistic frameworks are used for variant calling, then reliability is improved, but productivity deteriorates due to confounding factors
Solution Approach 1:
The patent dynamically adjusts key parameters in the probabilistic variant calling framework based on local sequence characteristics. For regions with highly related sequences, the patent modifies parameters such as minimum read depth thresholds, allele frequency cutoffs, and confidence score requirements to account for the increased complexity and potential confounding factors, thereby maintaining reliability while improving processing efficiency in these challenging regions.
Data Source
Figure 1
Figure 2
AI summary
Systems and methods for in silico prediction of HLA type of a patient are presented in which patient sequence reads and a reference sequence with known and distinct HLA alleles are used in a de Bruijn graph. A composite match score is then used to rank HLA alleles, thus providing a first HLA type. A second HLA type is identified by re-ranking using an adjusted composite match score.