Structural Variant Detection via Barcode Overlap in Targeted Sequencing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing algorithms for detecting structural variants in whole genome sequencing data are not applicable to targeted sequencing data, such as whole exome sequencing, due to the inability to span breakpoints, which are often located in gene introns rather than exons, making it difficult to detect gene fusions in cancer and other genetic disorders.
Innovation Solution
A computing system and method using an expectation-maximization approach to determine the integrity of query strings with respect to a ground truth string by fragmenting and partitioning the genome, employing barcode information to identify structural variation breakpoints through significant barcode overlap, and iteratively adjusting models to converge on the integrity of the strings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional breakpoint detection algorithms are used, then detection is possible when breakpoints are spanned by read pairs, but detection fails when breakpoints are not spanned (e.g., in gene introns)
Solution Approach 1:
The genome is divided into discrete bins or regions, and barcode information is aggregated at the bin level rather than requiring individual read pairs to span breakpoints. This segmentation allows detection of structural variants even when no single read pair spans the breakpoint, as the barcode overlap signal accumulates across multiple reads mapping to adjacent bins.
Solution Approach 2:
Barcode information serves as an intermediary signal that bridges the gap between reads that do not directly span breakpoints. By using barcodes as a mediator, the system can infer breakpoint locations through significant barcode overlap between adjacent bins, even when the breakpoints themselves are not directly observed in any single read pair.
2Measurement precision
If read pairs must span breakpoints for detection, then detection accuracy is high for spanned breakpoints, but detection sensitivity is low for non-spanned breakpoints
Solution Approach 1:
Multiple individual read signals are merged by aggregating barcode information across many reads mapping to genomic bins. This combining of weak individual signals into a collective barcode overlap signal enables detection of breakpoints with high sensitivity, while the bin-based approach maintains precision by localizing the signal to specific genomic regions.
Solution Approach 2:
The method uses excessive sampling of barcode information across many reads and bins to compensate for the fact that no single read pair may directly span the breakpoint. By accumulating partial signals from numerous reads, the system achieves both high sensitivity and maintained precision through the aggregate barcode overlap pattern.
Data Source
AI summary
Systems and methods for analyzing first and second strings against a ground truth string are provided. A construct representing a plurality of components is obtained, each component for a different portion of the truth string. The construct comprises a plurality of measurement string sampling pools each having an identifier and a corresponding plurality of measurement samplings corresponding to one or two of the components. Each sampling has the identifier and a portion of the first or second string. Samplings are assigned to first, second or third classes when coding a portion of the first string, second string, or both the first and second string. First and second positions are tested for sequence events by calculating a plurality of sequence event models using assumptions on the components having samplings encompassing the first and second positions and class assignments. These assumptions are updated using the calculated models and the models are recalculated.