Overlapping Reference Cell Arrays for Parallel Sample Read Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional genome sequencing methods face challenges in scalability, computational cost, and memory resource utilization due to the need to compare large amounts of data, particularly in de novo and reference-aligned sequencing, which limits the number of compute threads and cores that can be used.
Innovation Solution
A reference-guided device that preprocesses sample reads by using a system with arrays of cells storing overlapping reference sequences, allowing for probabilistic location of sample reads within a reference genome, thereby reducing memory requirements and computational costs, and improving scalability by enabling more efficient sorting and processing of data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the full reference genome is stored in shared memory for reference-aligned sequencing, then memory accessibility is improved, but the number of compute threads that can access the memory is limited
Solution Approach 1:
The reference genome is divided into multiple segments or partitions, with each compute thread assigned to process a specific segment. This segmentation allows multiple compute threads to access different portions of the reference genome simultaneously through shared memory, thereby increasing the number of concurrent compute threads while maintaining memory accessibility.
Solution Approach 2:
The patent introduces a new dimension of parallelism by organizing the reference genome access in a distributed manner across multiple compute threads, each handling specific regions. This transforms the single-dimension shared memory access model into a multi-dimensional parallel access model, enabling more compute threads to operate concurrently without overwhelming the shared memory bandwidth.
2Reliability
If sample reads are analyzed as a single large group for de novo sequencing, then comprehensive analysis is improved, but memory resources and computational cost increase significantly
Solution Approach 1:
The patent segments the large group of sample reads into smaller subgroups that can be processed independently by different compute threads. Each compute thread processes a specific subgroup of sample reads against its assigned reference genome segment, reducing the memory footprint required per thread while maintaining comprehensive analysis through coordinated processing of all subgroups.
Solution Approach 2:
Instead of processing all sample reads simultaneously against the entire reference genome, the patent applies partial action by having each compute thread process a subset of sample reads against a subset of the reference genome. This distributed partial processing reduces overall memory requirements while achieving complete coverage when results are aggregated.
3Productivity
If each compute thread is assigned a large dedicated memory for reference-aligned sequencing, then processing capability is improved, but scalability is limited
Solution Approach 1:
The patent merges the reference genome storage into shared memory that is accessible by all compute threads, eliminating the need for each thread to have a dedicated copy of the full reference genome. This combining of resources allows compute threads to share the memory burden, improving scalability while maintaining processing capability through efficient shared access patterns.
Solution Approach 2:
The shared memory system is designed to serve multiple compute threads simultaneously, making it a universal resource that supports scalable expansion. The shared memory infrastructure can accommodate an increasing number of compute threads without requiring proportional increases in total memory capacity, as each thread accesses only its assigned portion of the reference genome.
Data Source
AI summary
A device for locating a sample read with respect to a reference genome includes a plurality of groups of cells. Each group of cells stores a reference sequence representing reference bases from the reference genome corresponding to an order of cells in the respective group of cells. Each group of cells further stores a current substring sequence representing sample bases from the sample read corresponding to the order of the cells in the respective group of cells. Each group of cells stores the same current substring sequence and a reference sequence representing a portion of the reference genome that partially overlaps at least one other portion of the reference genome represented by one or more other reference sequences stored in one or more other groups of cells. Groups of cells are identified among the plurality of groups of cells where the stored reference sequence matches the current substring sequence.