Duplicate Marking in Gene Sequencing Pipelines With Linked-List Metadata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gene sequencing systems face challenges in terms of performance and storage resource utilization, particularly in the seeding sub-stage of the SR Alignment sub-system, due to inefficient memory access patterns in the Burrows Wheeler Transform (BWT) algorithm, which leads to prolonged processing times and high storage requirements.
Innovation Solution
The proposed solution involves optimizing memory access by storing frequently accessed occurrence table entries in a faster memory type, reducing the need for random and sequential memory accesses during the seeding sub-stage, and utilizing a hardware device to improve performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional gene sequencing systems process large datasets using traditional methods, then processing can be performed, but processing time extends to days and storage space consumption reaches hundreds of Gigabytes
Solution Approach 1:
The patent divides the occurrence table into multiple parts and stores them in different memory types (fast memory and slow memory). This segmentation allows frequently accessed data to be in fast memory while less frequently accessed data resides in slow memory, significantly reducing processing time for large datasets without requiring all data to be in fast memory simultaneously.
Solution Approach 2:
The patent introduces a hierarchical memory structure with multiple levels (fast memory and slow memory), adding a spatial dimension to memory organization. This multi-dimensional memory architecture enables efficient data access patterns by placing different portions of the occurrence table in different memory layers based on access frequency and timing requirements.
2Reliability
If conventional gene sequencing systems store complete occurrence tables for reference genome sequences, then alignment operations can be performed, but storage space consumption reaches hundreds of Gigabytes
Solution Approach 1:
The patent segments the occurrence table into multiple parts distributed across different memory types. This segmentation reduces the storage burden on any single memory system while maintaining the completeness and accuracy of the reference genome data needed for reliable alignment operations.
Solution Approach 2:
The patent applies different memory characteristics to different parts of the occurrence table based on local access patterns. Frequently accessed portions are stored in fast memory with high speed but limited capacity, while less frequently accessed portions are stored in slow memory with larger capacity but lower speed, optimizing both storage efficiency and alignment accuracy.
3Speed
If the occurrence table is stored entirely in fast memory, then memory access time is reduced, but the cost and size of fast memory becomes prohibitively large
Solution Approach 1:
The patent divides the occurrence table into segments that fit within the capacity of fast memory, storing only the most frequently accessed segments in fast memory while placing less frequently accessed segments in slow memory. This segmentation strategy optimizes memory access time for critical operations while keeping fast memory size within practical limits.
Solution Approach 2:
The patent implements a dynamic memory management strategy where the division of the occurrence table between fast and slow memory can be adjusted based on access patterns and available resources. This dynamic allocation allows the system to adapt to different workloads and optimize the balance between access speed and memory size requirements.
Data Source
AI summary
In accordance with embodiments, a processing unit performs alignment of a short read (SR) against a reference genome sequence. The processing unit determines whether the SR is aligned. If the SR is not aligned, the processing unit receives the next SR and processes the next SR by repeating. If the SR is aligned, in response to the determination that the SR is aligned with the reference genome sequence at a first position in the reference genome sequence, the processing unit generates a new SR metadata entry corresponding to the SR. The processing unit finds a linked list in a SR metadata collection. The first position of the linked list in the SR metadata collection corresponds to the first position of the reference genome sequence where the SR is aligned. The processing unit performs duplicate marking based on the SR and the linked list.