CUDA-EC

CUDA-EC corrects sequencing errors in high-throughput short-read DNA data to produce error-free reads for de novo genome assembly and downstream genomic analyses.

Key Features:

• Scalable parallel algorithm: Implements a scalable parallel algorithm using the Compute Unified Device Architecture (CUDA) programming model to perform error correction across large short-read data sets.
• Spectral alignment: Uses spectral alignment to identify and correct errors within sequencing reads by analyzing their spectral properties.
• CUDA texture memory utilization: Leverages CUDA texture memory to enhance computational efficiency during error correction processes.
• Space-efficient Bloom filter: Incorporates a space-efficient Bloom filter data structure for spectrum membership queries.

Scientific Applications:

• Graph-based short-read assembly support: Provides corrected reads to graph-based short-read assembly tools to improve the accuracy of de novo genome assembly.
• Illumina sequencing data processing: Applicable to real and simulated Illumina sequencing data across varying read lengths, error rates, and input sizes.

Methodology:

Implements a scalable parallel algorithm using the CUDA programming model and spectral alignment; uses CUDA texture memory and a space-efficient Bloom filter for spectrum membership queries; tested on real and simulated Illumina data sets and reported speedups of 12-84× for parallelized error correction and 3-63× versus the Euler-SR program for sequential preprocessing and parallelized error correction.