De Novo Nucleotide Assembly Using Modified String Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gene assembly systems face challenges in efficiently processing large nucleotide sequence datasets due to high computational overhead and long processing times, particularly when dealing with enormous data sets that exceed the capacity of volatile memory.
Innovation Solution
The layout assembly system employs a modular approach with a mer generating and sorting system, hit determination system, hit table sorting system, and post processing system to efficiently compare and align nucleotide sequences by dividing sequences into mers, sorting, and condensing hits, allowing for rapid processing of large datasets without the need for loading entire sequences into volatile memory.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional gene assembly systems process entire datasets in a single operation, then complete genome assembly is achieved, but processing time becomes excessively long and memory requirements exceed volatile memory capacity
Solution Approach 1:
The patent divides the genome assembly process into distinct phases: preprocessing reads into smaller units, sorting them into buckets, performing pairwise comparisons only within buckets, and progressively merging results. This segmentation allows processing of large datasets in manageable chunks rather than requiring entire datasets in memory simultaneously, thus reducing processing time while maintaining assembly completeness
Solution Approach 2:
The patent performs preliminary sorting and bucketing of sequence reads before the main assembly process. By pre-organizing data into buckets based on sorting criteria (such as k-mer content or genomic position estimates), the system prepares data structures that enable faster subsequent comparisons, avoiding the need for exhaustive pairwise comparisons of all reads against all other reads
2Ease of operation
If conventional systems load entire sequence datasets into volatile memory, then random access processing is enabled, but memory requirements exceed available volatile memory capacity
Solution Approach 1:
The patent segments the sequence dataset into multiple buckets that can be processed independently. Each bucket contains a subset of reads that can be loaded into volatile memory, processed, and then merged with other bucket results. This segmentation enables efficient use of limited volatile memory while still allowing random access within each bucket during processing
Solution Approach 2:
The patent introduces bucket structures as intermediary data containers between the raw sequence dataset and the final assembly. These buckets act as intermediate storage units that can be efficiently manipulated in volatile memory, serving as a mediator that reduces the memory footprint required for processing while maintaining data accessibility
3Measurement precision
If exhaustive pairwise comparison of all sequences is performed, then complete alignment coverage is achieved, but computational overhead increases dramatically
Solution Approach 1:
The patent segments the comparison space by organizing reads into buckets where comparisons are performed only within each bucket rather than across all reads. This segmentation maintains alignment accuracy for sequences that are genomically proximal while dramatically reducing the total number of comparisons required, thus lowering computational overhead
Solution Approach 2:
The patent applies different processing strategies to different local regions (buckets) of the genome. By performing localized assembly within each bucket and then merging results, the system achieves accurate alignment for each local region while avoiding the computational burden of global pairwise comparison, effectively applying local quality optimization
4Productivity
If raster-type scan organization is used for pairwise comparisons, then systematic processing is achieved, but computational overhead becomes prohibitive for large datasets
Solution Approach 1:
The patent replaces the traditional raster-type scan with a bucket-based segmented processing approach. Instead of systematically comparing reads in sequential order across the entire dataset, the system divides comparisons into independent bucket operations, enabling parallel processing and reducing the computational overhead associated with sequential raster scanning of large datasets
Data Source
AI summary
Systems and methods to automatically de novo assemble a set of unordered read sequences into one or more, larger nucleotide sequences are presented. The method involves first creating two identical sets of the reads, dividing each read in both sets into smaller sorted mer sequences and then comparing the mers for each read in set 1 to the mers from each read in set 2 to exhaustively identify overlapping segments. Overlap information is used to construct a modified assembly string graph, traversal of which produces a sorted string graph layout file consisting of all the reads ordered left to right including their approximate starting offset position. The sorted string graph layout file is then processed by a novel multiple sequence alignment system that uses mer matches between all the overlapping reads at a given position to place matching individual bases from each read into columns from which an overall consensus sequence is determined.