Alignment-Free Fusion Detection Using Kmer Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting fusion events from sequencing data, particularly in cell-free nucleic acid samples, face challenges such as high computational time and memory usage due to large data volumes and low cancer allele fractions, leading to lower than 75% true positive rates and vulnerability to false positives.
Innovation Solution
The method involves processing sequencing data into fragments, decomposing them into kmers, and analyzing these kmers to identify gene ranges that span across fragments, allowing for the detection of candidate fusion events without relying on alignment, thereby reducing computational resources and improving speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If alignment-based fusion detection methods are used on cell-free nucleic acid data, then fusion events can be identified, but computational time and memory usage become debilitating due to large data volumes
Solution Approach 1:
The patent segments the sequencing data processing into distinct phases: initial alignment to identify candidate fusion reads, followed by assembly of these reads into contigs, and finally verification of fusion events. This segmentation allows the computationally intensive alignment step to be performed only on a subset of data that is most likely to contain fusion events, rather than aligning all sequencing reads, thereby reducing overall computational time while maintaining detection accuracy
Solution Approach 2:
The patent extracts and focuses specifically on discordant reads that show alignment patterns consistent with fusion events, separating these from the vast majority of reads that do not contain fusion information. By extracting only the relevant subset of reads for further processing, the method reduces the computational burden while preserving the ability to detect fusion events accurately
2Measurement precision
If alignment-based fusion detection methods are used on cell-free nucleic acid data, then fusion events can be identified, but memory consumption becomes excessive due to large data volumes
Solution Approach 1:
The processing pipeline is divided into stages where data is handled in manageable chunks. After initial alignment, only discordant reads are retained and assembled into contigs, dramatically reducing the data volume that needs to be held in memory simultaneously. This staged approach allows accurate fusion detection while keeping memory consumption within practical limits
Solution Approach 2:
The method extracts only the essential information needed for fusion detection - specifically the discordant reads and their alignment characteristics - and discards or processes the remaining data more efficiently. This extraction of critical information reduces the memory footprint while maintaining detection sensitivity
3Measurement precision
If existing fusion callers push sensitivity higher to detect low cancer allele fractions, then detection sensitivity improves, but false positive discoveries increase
Solution Approach 1:
The patent implements a feedback mechanism where assembled contigs are verified against multiple criteria including alignment consistency, read support thresholds, and biological plausibility checks. This feedback loop allows the method to maintain high sensitivity for detecting low-abundance fusion events while filtering out false positives through iterative validation, thereby improving both sensitivity and reliability simultaneously
Solution Approach 2:
The method performs preliminary assembly of discordant reads into contigs before final fusion event calling, creating a more robust signal that is less susceptible to false positives. This preliminary action of consolidating supporting reads into contigs provides a stronger basis for confident fusion detection, enabling high sensitivity while maintaining reliability
4Measurement precision
If misaligned fusion supporting reads are recovered through assembly, then reading inaccuracies are reduced, but computational complexity increases
Solution Approach 1:
The assembly process is applied only to discordant reads that are candidates for fusion events, rather than attempting to assemble all sequencing reads. This segmentation limits the computational complexity of the assembly step to a manageable subset of data, while still recovering misaligned reads that contain fusion information and improving reading accuracy for these critical events
Data Source
AI summary
Cell free nucleic acids from a test sample obtained from an individual are analyzed to identify possible fusion events. Cell free nucleic acids are sequenced and processed to generate fragments. Fragments are decomposed into kmers and the kmers are either analyzed de novo or compared to targeted nucleic acid sequences that are known to be associated with fusion gene pairs of interest. Thus, kmers that may have originated from a fusion event can be identified. These kmers are consolidated to generate gene ranges from various genes that match sequences in the fragment. A candidate fusion event can be called given the spanning of one or more gene ranges across the fragment.