Graph-Based Local Assembly for Reference-Bias-Free Variant Calling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Next-generation sequencing (NGS) methods face challenges in re-assembling sequence reads into a consensus sequence that are computationally expensive and require alignment to a reference sequence, especially when dealing with divergent genomes.
Innovation Solution
A method involving the creation of a directed acyclic graph (DAG) from nucleotide sequence reads, where sequence reads are aligned to each other and the graph, allowing for variant calls without initial reference alignment, and incorporating a graph-aware alignment algorithm like Smith-Waterman or Graph-Smith-Waterman to build a detailed representation of the genomic region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional alignment algorithms are used to assemble sequence reads into a consensus representation, then the alignment accuracy is improved, but the computational expense increases significantly
Solution Approach 1:
The patent divides the genome into smaller genomic regions of interest and performs local assembly on each region separately. This segmentation approach allows the use of computationally intensive alignment algorithms only on small, targeted regions rather than the entire genome, thereby maintaining alignment accuracy while significantly reducing overall computational expense.
Solution Approach 2:
The patent applies different assembly strategies to different genomic regions based on their specific characteristics. For regions with high variability or repetitive sequences, more sophisticated local assembly methods are employed, while simpler methods are used for stable, non-repetitive regions. This local quality approach optimizes computational resource allocation while maintaining necessary accuracy where needed.
2Ease of manufacture
If short sequence reads from NGS are aligned to a reference sequence, then the assembly process is simplified, but the alignment quality deteriorates due to short read lengths
Solution Approach 1:
The patent segments the assembly process into two stages: first, short reads are aligned to identify regions of interest; second, local assembly is performed on these specific regions using the short reads aligned to each other and to the reference. This segmentation allows the short reads to be used effectively for their intended purpose while achieving high-quality assembly in targeted regions.
Solution Approach 2:
The patent uses the reference sequence as an intermediary to guide the local assembly process. The reference sequence helps identify regions of interest and provides a framework for assembling short reads, thereby compensating for the limitations of short read lengths while maintaining assembly simplicity.
3Adaptability or versatility
If de novo assembly is performed without reference alignment, then reference bias is eliminated, but the computational complexity increases
Solution Approach 1:
The patent performs de novo local assembly on segmented genomic regions rather than on the entire genome. This segmentation reduces the computational complexity of reference-independent assembly by limiting the scope to small, manageable regions while still eliminating reference bias in those regions.
Solution Approach 2:
The patent applies de novo assembly selectively to specific genomic regions of interest rather than performing it genome-wide. This partial action approach eliminates reference bias where needed while avoiding the excessive computational complexity that would result from genome-wide de novo assembly.
Data Source
AI summary
Computer-implemented methods and systems for performing a local assembly of a genomic region of interest include the de novo or assisted creation of a directed graph, such as a directed acyclic graph (DAG), from a plurality of obtained nucleotide sequence reads. First and second sequence reads are aligned to each other to define at least one node of the DAG. Successive alignments of the remaining sequence reads to the then-defined DAG are performed to extend nodes and/or add nodes to the DAG. Graph-aware alignment techniques that produce alignment scores or indicators are employed in defining the nodes of the DAG from the sequence reads. The created DAG represents and describes in detail the genomic region of interest and can be used to perform variant calls.