ParLECH

ParLECH corrects indel and substitution errors in PacBio long reads by leveraging Illumina short-read k-mer coverage and de Bruijn graph information to improve sequencing accuracy.

Key Features:

• Hybrid correction: Integrates Illumina short reads with PacBio long reads to perform hybrid error correction.
• De Bruijn graph construction: Builds a de Bruijn graph from Illumina short reads to provide a reference structure for correction.
• Indel correction via widest path: Replaces indel errors by the widest path (maximum min-coverage path) in the short-read de Bruijn graph.
• K-mer coverage segmentation: Uses k-mer coverage information to segment each long read into low- and high-coverage regions.
• Substitution correction by majority voting: Applies a majority voting mechanism within segmented regions to correct substitution errors.
• Error-region identification: Employs k-mer coverage to distinguish erroneous from accurate sequence regions.
• Distributed, scalable algorithm: Implements a parallel algorithm capable of scaling to hundreds of compute nodes.
• Large-scale processing: Designed to handle terabyte-scale sequencing datasets for high-throughput correction tasks.

Scientific Applications:

• Human genome correction: Corrected indel errors on a 312 GB PacBio long-read human dataset using 452 GB of Illumina short reads in under 29 hours on 128 compute nodes.
• Microbial genomics (E. coli): Achieved alignment of over 92% of bases from an E. coli PacBio dataset to its reference genome.
• High-throughput sequencing projects: Applicable to large-scale projects requiring correction of long reads across terabytes of data using hundreds of nodes.

Methodology:

Construct a de Bruijn graph from Illumina short reads; identify and correct indels by replacing them with the widest (maximum min-coverage) path in that graph; segment long reads into low/high coverage regions using k-mer coverage; correct substitutions by majority voting within segments; implemented as a distributed, scalable algorithm.