BLESS: NGS Error Correction Tool Wiki

BLESS: Bloom-filter-based Error Correction Solution for High-throughput Sequencing Reads

Developed by ESCAD Group, Computational Comparative Genomics Lab, and IMPACT Group
in Univsersity of Illinois at Urbana-Champaign

Change History

V0.11 10/24/2013
First release.

V0.12: 12/02/2013
Single-end (or merged paired-end) reads are supported.
A bug in correcting errors in the first k-mer is corrected.

V0.13: 04/06/2014
Error correction accuracy has been improved especially for highly repetitive genomes.
BLESS supports dumping/loading the Bloom filter contents.
Reads are trimmed by default.

V0.14: 04/08/2014
Everything is equal to V0.13 except that this version supports Illumina 1.8+ FASTQ files.

V0.15: 05/16/2014
A bug that may occurs when overall quality scores are very bad is fixed.
Reads that have more than MAX_N_RATIO * <read length=""> Ns are modified (default MAX_N_RATIO: 0.1).</read>

V0.16: 05/24/2014
Memory leak that may happen in a read with many low-quality-score bases is fixed.

V0.17: 06/03/2014
A bug in setting the maximum allowed Ns for paired reads is fixed.

V0.20: 10/16/2014 (Preliminary release of BLESS 2)
BLESS is parallelized using OpenMP and MPI.
Reads with different length are supported.
The k-mer counting part is substituted with KMC (http://sun.aei.polsl.pl/kmc).
Even numbers can be used for k.
Some minor bugs are fixed.

V0.21: 12/06/2014
White space characters in correted read files are removed.
The multi-threaded runtime is improved.

V0.22: 12/17/2014
MPI is applied to KMC.

V0.23: 12/19/2014
Some bugs are fixed.

V0.24: 01/29/2015
KMC is updated to 2.2.1 (The max memory usage of KMC can be fixed to 4 GB for any genome).
A bug in writing a quality score hisgoram file is fixed.

V1.00: 05/09/2015 (aka BLESS 2)
Error correction algorithm is enhanced.
Compressed input/output files are supported.
Dependent libraries are included (Boost headers are still needed).
Non-ACGTN characters can be handled.
Non deterministic behaviors are corrected.

V1.01: 05/10/2015
Runtime is improved by 10%.
All the dependent libraries are included.
Errors in compiling pigz are fixed.

V1.02: 06/10/2015
A bug in single-end reads with many Ns is fixed.

System Requirements

Compiling BLESS requires MPI libraries. The latest version of BLESS was tested
with GCC 4.9.2, MPICH 3.1.3 (or OpenMPI 1.8.2).
If you see the following error message, you need to upgrade your MPI library:
error: initializing argument 2 of 'int MPI_File_open(MPI_Comm, char, int, MPI_Info, ompi_file_t)'*

Dependency

MPI: tested using MPICH 3.1.3 and OpenMPI 1.8.4

(Included dependent libraries)
Boost library (http://www.boost.org)
Google sparsehash (https://code.google.com/p/sparsehash)
klib (https://github.com/attractivechaos/klib)
KMC (http://sun.aei.polsl.pl/REFRESH/index.php?page=projects&project=kmc&subpage=about)
murmurhash3 (https://code.google.com/p/smhasher)
zlib (http://zlib.net)
pigz (http://zlib.net/pigz)

How to Install BLESS

Type make.

How to Run BLESS

Single-end reads or merged paired-end reads
> ./bless -read <fastq> -prefix <output prefix> -kmerlength <k-mer length>
<fastq> : fastq file name
<output prefix>: <output directory name>/<file prefix>
<k-mer length> : k-mer length (odd number)

Paired-end reads
> ./bless -read1 <forward fastq> -read2 <reverse fastq> -prefix <output prefix> -kmerlength <k-mer length>
<forward fastq>: first read file of a paired-end fastq file
<reverse fastq>: second read file of a paired-end fastq file
<output prefix>: <output directory name>/<file prefix>
<k-mer length> : k-mer length (odd number)

Run bless with no option to see the entire options.

Parallelizing BLESS

BLESS is parallelized using OpenMP and MPI. The number of threads per each node can be changed
using "-smpthread" option (default: the number of cores in a SMP node).

Using multiple nodes using MPI
Launch single process on a node and bind the processes to their nodes
Ex) MPICH 3.1.3 Hydra Version
> mpirun -ppn 1 -bind-to board ./bless -read1 <forward fastq> -read2 <reverse fastq> -prefix <output prefix> -kmerlength <k-mer length>

Choosing k-mer Length

Our empirical analysis shows that the k value that satisfies the following two conditions usually generates the results close to the best one:
1) Ns / 4^k ≤ 0.0001 where Ns represents the number of unique solid k-mers (BLESS reports Ns).
2) Number of corrected bases becomes the maximum at the chosen k value, which means you need to increase k as long as the number of corrected errors increases.

How to Load Existing Bloom Filter Data

If you run BLESS with the "-prefix prefix" option, BLESS automatically generates the files prefix.bf.data and prefix.bf.size that have the contents of the Bloom filter.
These files can be loaded into BLESS to skip time-consuming the Bloom filter programming process.

Single-end reads or merged paired-end reads
> ./bless -read <fastq> -load prefix -prefix <new prefix> -kmerlength <k-mer length>

Paired-end reads
> ./bless -read1 <forward fastq> -read2 <reverse fastq> -load prefix -prefix <new prefix> -kmerlength <k-mer length>

Publications

Yun Heo, Xiao-Long Wu, Deming Chen, Jian Ma, Wen-Mei Hwu (2014). BLESS: Bloom filter-based error correction solution for high-throughput sequencing reads. Bioinformatics, 30 (10), 1354-1362. doi: 10.1093/bioinformatics/btu030

Yun Heo, Anand Ramachandran, Wen-Mei Hwu, Jian Ma, Deming Chen (2016). BLESS 2: accurate, memory-efficient and fast error correction method. Bioinformatics, 32 (15): 2369-2371. 10.1093/bioinformatics/btw146

Bug Reports

Anand Ramachandran <aramach4 at illinois.edu>